Comparison of Rehabilitative Interventions That Ameliorate Post-stroke Working Memory Deficit: A Systematic Review

Stroke is one of the most common causes of disability in the world. It has sensory, motor, and cognitive symptoms. Many cognitive domains might get involved in a stroke. This systematic review focuses on working memory domain deficits after stroke and their various rehabilitation methods. This review is based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses(PRISMA) guidelines. For this review, we have searched PubMed, Google Scholar, and Science Direct databases and screened thoroughly with the inclusion criteria of free full-text English papers in the last 10 years that have exclusively studied humans. The articles included in the search are randomized control trials (RCTs), observational studies, meta-analysis studies, systematic reviews, and traditional reviews. Consequent quality assessment was done using the most commonly used tools for each type of study and eight papers were selected. From these papers, full-text articles were studied, analyzed, and tabulated. We found five different rehabilitation methods: transcranial direct-current stimulation, computer-assisted cognitive rehabilitation, physical activity, goal setting, and multimodal rehabilitation. We found that goal setting, computer-assisted cognitive rehabilitation, and multimodal rehabilitation can improve working memory deficits. While transcranial direct current stimulation and physical activity were inconsistent, further studies are needed. The small sample size, no follow-up, the inclusion of only a few studies, the size of the stroke, and comorbid conditions like mild cognitive impairment, dementia, and depression were the main limitations of this study. Future reviews must include a larger number of studies with large sample sizes, including follow-up as an inclusion criterion. We need more clinical trials on these methods for better knowledge.


Introduction And Background
Stroke is the second most common cause of dementia and the third most common cause of morbidity [1]. It can occur due to ischemic (most common) or hemorrhagic causes and can decrease the blood supply of the brain. The most common involved symptoms besides physical deficits (such as complete or partial paralysis, sensory loss, and altered sensations) are dysfunction in learning, memory, and executive functions. This affects nearly 83% of post-stroke patients with cognitive dysfunction, and more than 60% of stroke survivors report cognitive dysfunction, even for up to 10 years [2]. While motor and sensory deficits can affect patients' quality of life and economic status, memory deficit will affect patients' daily activities, their profession, and their families. Although the motor and cognitive functions are fundamentally treated as well-defined and separate entities while treating and diagnosing, both entities play an essential role in the post-stroke assessment of behavior and disability [3]. When it comes to cognitive function, working memory has a fundamental role in performing complex behavior [4]. Damage to working memory function causes a drop in the complex cognitive function of the brain to perform everyday activities such as memorization, communication, planning, reading, and writing [5].
What is working memory? It is defined as a multi-component system involved in goal-directed behavior that involves retaining and manipulating information [6]. In simple terms, it is explained as a "sketchpad of conscious thought" [7]. for example, problem-solving in our mind and navigating to our home within our mind. Baddeley and Hitch described it as an essential model with four sub-components arranged hierarchically [6]. They are the phonological loop, visuospatial sketchpad, central executive, and episodic buffer [6]. The central executive supervises the other three subcomponents to store visual, spatial, and phonological information [8]. Working memory reaches its optimal capacity by 20 to 25 years of age. Then each subcomponent declines slowly with further aging at a different rate, which is explained by visual memory declining faster than phonological memory [9]. Many previous imaging studies found that working memory is a part of the lateral prefrontal and parietal cortex of the brain [10]. These areas are supplied by the middle and anterior cerebral artery branches and later drained into cerebral venous sinuses. The stroke affecting these regions will impair the functions.
Though post-stroke dementia, episodic memory, and long-term memory deficits after stroke have been studied abundantly, working memory deficits after stroke have less information. There is ongoing research on various methods to improve working memory skills and reduce the patient's morbidity on the efficacy of each rehabilitation. To our knowledge, there is a lack of information on the comparison and efficacies of each intervention; such information is vital for selecting an effective intervention. This review article compares rehabilitative interventions to reduce the working memory deficit after stroke.

Review Methods
This systematic review was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [11].

Search Strategies
We performed this systematic review using PubMed, Google Scholar, and Science Direct databases with keywords "stroke" and "working memory deficits" in all three databases, including Medical Subject Headings (MeSH) terms and keywords in PubMed search, and included multiple filters for each journal. These details are illustrated in Table 1.

Inclusion and Exclusion Criteria
We have included free, full-text English papers that have exclusively studied humans in the last 10 years. The articles included are randomized control trials (RCTs), observational studies, meta-analysis studies, systematic reviews, and reviews.

Selection Strategy
Two reviewers selected the articles independently using the same search strategy in all three databases. At first, articles were screened from the title of articles and abstracts and then later by reading full-text articles. If contradicting results regarding the article's eligibility occurred, reviewers assessed the full-text article until the group reached a consensus.

Data Collection, Items, Analyses, and Outcome Assessment
Two reviewers collected data independently, and if reviewers found contradicting results, a full-text article was scrutinized and discussed until all reviewers reached a consensus. Collected data were analyzed and tabulated under various headings such as (i) first author and year, (ii) population along with dropout patients (for RCTs)/number of studies included (for systematic review, traditional review, and metaanalysis, (iii) intervention, (iv) duration of intervention, (v) outcome measures and assessment, (vi) followup assessment, and (vii) funding sources. Due to heterogeneity in the population, different interventions, and unique outcomes of each study when compared with other studies, it was not possible to perform the meta-analysis.

Risk of Bias Assessment
Each selected study was assessed for risk of bias by two reviewers independently using commonly used tools for each type of study, and only studies that scored more significant than 70% were included in this review. Table 2 shows the quality assessment tools used for each type of study and accepted studies. Fishman et al. [12]. Studer et al. [13]. Park et al. [14].

Results
There were 174 potentially related titles found in the database search. One hundred and fifty-nine records were kept after duplicates were removed. When the titles and abstracts of these records were evaluated based on the qualifying criteria for this review, 25 articles remained. From these, 13 reports were discarded because of irrelevant data. Eight papers with a score of more than 70% were allowed in the review after each publication underwent a quality assessment. There were four RCTs, one traditional review, one metaanalysis, one systematic review, and one pilot research. A flow diagram of the study selection and screening process is shown in Figure 1.

FIGURE 1: Flowchart of the study selection process
CCRBT: Cochrane collaboration risk-of-bias tool

Study Characteristics
The main characteristics of the clinical trials, pilot study, meta-analysis, traditional review, and systematic review are shown in Tables 3, 4 respectively. Of these eight studies, one pilot study and one traditional review used transcranial direct current stimulation intervention, one systematic review and the same pilot study used computer-assisted cognitive rehabilitation (CACR) intervention, one RCT and one meta-analysis focussed on physical activity as intervention, one RCT applied goal setting as intervention, and one RCT adopted multimodal rehabilitation with follow-up. All these articles assessed the working memory domain using different outcome measures.

Outcome Measures
Working memory is usually measured using the digit span, letter-number sequencing, Wechsler spatial span, and visual cognitive performance tests. Each article used one or more of these tests. Though these tests help assess improvement in working memory, much more ecologically valid measures are required to evaluate real-life situations.

Discussion
In this systematic review, we aimed to compare the efficacies of rehabilitation that reduce working memory deficits after stroke. We selected five randomized control trials, two systematic reviews, and one traditional review and collectively showed five different rehabilitation methods to improve working memory. For better understanding, all included papers were divided into rehabilitation methods and discussed in detail.

Transcranial Direct Current Stimulation (tDCS)
In our review, two papers were included regarding transcranial direct stimulation of the brain. Although tDCS has the advantage of being a non-invasive procedure that delivers a low-intensity direct current through the scalp, it still has areas to hone itself because of the presence of variables from person to person, such as conductivity, the resistance of the scalp, and the dosage of current. Though there are many types of tDCS, Anodal tDCS is the most commonly used and it has significant evidence from many publications that enhance working memory when stimulated by the dorsolateral prefrontal cortex (DLPFC) [18].
In healthy individuals, anodal tDCS has increased correct responses and fewer errors were noted when compared to other types of tDCS such as cathodal or sham tDCS [18]. Some studies described that anodal tDCS promotes neural plasticity in chronic stroke survivors, which were shown as alpha waves in the electroencephalography (EEG) [19]. These alpha waves indicate functional activity of working memory along with attention, motor learning, and performance and are considered a biomarker of functional activity of the brain [19,20]. Anodal tDCS application is based on the inter-hemispheric imbalance of excitability, which is explained by hyperexcited non-lesioned DLPFC on one hemisphere of the brain because of the loss of transcallosal inhibition from lesioned DLPFC on another hemisphere of the brain [18,20]. This imbalance in excitability is a poor functional recovery marker and is corrected by applying tDCS either with anodal stimulation, which causes an excitatory stimulation on the affected hemisphere, or cathodal stimulation, which causes an inhibitory stimulation on the non-affected hemisphere and brings back the balance in both hemispheres, which enhances recovery [18]. One pilot study describes that there is a synergistic effect when tDCS combined with CACR has shown short-term improvement in the verbal two-back working memory test [14]. But, due to the small number of patients (n=11) within the trial, this study cannot be generalized to the population, and there is no evidence of any improvements in tDCS in long-term rehabilitation.

Computer-Assisted Rehabilitation
In our review, a systematic review of computer-based cognitive training included four studies only on working memory improvement after computer-based rehabilitation training using Cogmed training (Cogmed, Stockholm, Sweden) either at a rehabilitation center (three studies) or at home (one study) [17]. It is a computer program that includes audio and visuospatial tasks and always requires a motor response. Patients were trained for 30 to 40 min for five days a week for a five-week duration [17]. This training improved working memory even after three months of rehabilitation [17]. Along with working memory, objective improvements such as attention, and subjective improvements such as mood were seen to improve [17]. This systematic review also included seven studies on combined training of working memory and attention, which showed improvement in all seven studies. Still, these results are based on one or two tasks.
In contrast to only working memory studies, this combined training showed only immediate recall, and only one out of seven studies showed improvement in mood. Though this is an extensively reviewed article, the lack of a proper control group and the lack of long-term evaluation of this training are the main limitations that should be addressed in future studies [17].
An RCT of 83 patients with the help of a computer game called Wizard memory game (Peak, Manchester, United Kingdom) showed a significant improvement in working memory [13]. It is a game with working memory tasks, and patients were divided into a pre-commitment group and a control group. The precommitment group was given a choice restricting option of visitor ban or continuous physician surveillance to ensure voluntary modification of one's choice to enhance the likelihood of reaching the target (decisional neuroscience) [13]. Decisional neuroscience encourages the patients to reset their behavior and promotes motivation to change behavior [13]. Precommitment and control groups are given a different frequency of training every second and fifth day of self-directed training for two weeks [13]. When comparing pre and post-changes in both verbal and visuospatial working memory, the statistical analyses have a positive correlation with the Wizard memory game. More improvements in working memory were found when this rehabilitation was used as an add-on to standard post-stroke neurorehabilitation. Thus computer-assisted decisional neuroscientific intervention promotes working memory improvement in patients when given as an add-on to standard therapy. Though this study has a proper sample size, adequate power, and randomization, follow-up studies are necessary for assessing working memory in everyday life and the sustainability of training.

Physical Activity
A six-month randomized control trial of 28 stroke survivors who had an episode ≥ one-year onset and had completed their rehabilitation was divided into an intervention group and a control group. While the intervention group received two sessions of exercise training and one session of recreation and leisure activity, the control group received usual care. The intervention group was found to have a 43% improvement in working memory [15]. The effect is dependent on the duration and type of training (aerobic or multimodal exercise training) involved [15,21]. Although very few studies had shown improvement in working memory [15], many studies had contradicting results to support this statement [1,2]. Some studies have shown improvement when combined cognitive and exercise interventions are included [22]. Because of the small sample size and other contradicting studies, further extensive sample studies are needed.
Unlike the single RCT, we also included a meta-analysis of randomized control trials for the effect of physical activity on post-stroke working memory deficits [2]. This meta-analysis included around 14 studies, but only five out of 14 studies had working memory as an outcome [2]. This study did not reach a statistical significance on working memory to support physical activity as an intervention [2].
Both studies included in this review have a small sample, not distributed evenly, worked only in the chronic phase after stroke, and one study has not shown statistical significance for the working memory domain [15,2]. Though this might not help us decide with the present studies, other cognitive domains were included along with working memory to study in these studies. Along with physical activity, domains such as selective attention, set-shifting, processing speed measures, conflict resolution, and mood are included in their outcome. Some of these outcomes have positive effects with statistical significance. Physical activities such as aerobic and recreational activities play a significant role in the development of these cognitive domains.

Goal Setting
National Clinical Guidelines promote goal setting as an essential aspect of stroke rehabilitation. Many studies have shown improvement in patients' confidence and motivation to work on rehabilitation with prior goal setting to improve motor and cognitive skills [23,24]. There are some different goal perspectives between patients and hospital staff that impede the development of the patient. The patient goal is usually long-term oriented, returning to normal and vaguely described. In contrast, the hospital staff goal is shortterm, specifically focused on impairment, and motivated by financial and organizational pressures [23]. Terminating these barriers with open discussion of the goal with the patient and their families and positively encouraging and promoting the environment created by staff enabled many patients to participate in the rehabilitation [23] actively. There are tools used for goal setting, such as Return to Work (RTW) goal working sheet by American Stroke Association and goal management training [12], and can also be created and tailored per patients' requirements. Integrating well-defined, achievable goals into the goal and working on them daily will motivate the patient to work on them. We included a RCT of goal setting and its effect on stroke survivors with accurate power and a good sample size. This trial showed improvement in the intervention group [12]. It is not attributable to age, education, time since stroke, sleep, depression, or vascular risk factors after multiple corrections using SPSS statistical software (IBM Corp., Armonk, New York, United States). This signifies the importance of 'goal setting' as a tool that covers a broad spectrum of various stroke survivors. While the trial has given promising results in rehabilitation, it did not have a follow-up and did not include aphasia and dementia patients. Further studies could provide a better opinion when including a follow-up and inclusion of aphasia and dementia patients.

Multimodal Rehabilitation
Multimodal rehabilitations promote patients to have a wide range of functions to help with motor, sensory and cognitive functions. Because of multiple different approaches simultaneously, these can have an additive or synergistic effect. Also, some studies show an enriched environment promotes recovery with better results than being alone and socially inactive. A RCT was included in our studies, which had one control group and two intervention groups in which one had rhythm and music therapy (R-MT) and the other had dance and horse riding therapy (H-RT) for 12 weeks with two sessions weekly [16]. Many outcomes, including working memory, were statistically analyzed. It was found that the R-MT group improved working memory, even after six months of rehabilitation. Other groups did not have statistical significance to provide a proper result. Unlike all the prior rehabilitation methods, this had a good follow-up and promising results that show the retention of improvements even after rehabilitation. This study had a small strength of participants from a single region. So, a broader study region with a better number of participants can help us better understand how to incorporate a multimodal approach into clinical settings.

Limitations
This study has several limitations such as the size of the stroke and comorbid conditions like mild cognitive impairment, dementia, and depression. Along with these factors, other major limitations are most of the studies included has a small sample size, no follow-up, and only mentioned their use in the chronic phase of stroke, so they cannot be generalized to the overall population. Some of these methods have not yet been thoroughly studied. Hence future studies on these methods with RCTs would benefit us to know the benefits and side effects of these methods.

Conclusions
There are several individual studies on each cognitive rehabilitation method, but to our knowledge, we have not found any study on the comparison of one with another rehabilitation method. We found five different rehabilitation methods in these studies, in which goal setting, CACR, and multimodal rehabilitation can improve working memory deficits. tDCS and physical activity were inconsistent and further studies are needed. From this review, the multifaceted approach with synergistic effects is seen to be the best approach because one method's disadvantage can be counterbalanced with another rehabilitation method. This systematic review has a few limitationsa, such as small number of studies included, small sample size in a few papers, and availability of follow-up only in two studies. Future studies must include large sample sizes and follow-ups as inclusion criteria. More RCTs are needed on these methods for better knowledge.

Conflicts of interest:
In compliance with the ICMJE uniform disclosure form, all authors declare the following: Payment/services info: All authors have declared that no financial support was received from any organization for the submitted work. Financial relationships: All authors have declared that they have no financial relationships at present or within the previous three years with any organizations that might have an interest in the submitted work. Other relationships: All authors have declared that there are no