Abstract
This study found that patients with different stroke severity may have different priorities that influence their self-perceived levels of overall recovery.
Patient-reported outcome (PRO) refers to outcome information that is directly reported by patients themselves (Weldring & Smith, 2013). PRO has been recommended as a critical outcome measure in health care, because it contributes information from patients’ perspectives that cannot be obtained using objective assessments (Katzan et al., 2017; Kingsley & Patel, 2017; Weldring & Smith, 2013). Assessments of PRO serve clinical and research purposes (Churruca et al., 2021; Price-Haywood et al., 2019). Clinically, PRO assessments help clinicians detect patients’ needs to provide corresponding interventions and monitor the outcome. In studies, PRO assessments aid in the evaluation of the efficacy of interventions and enable researchers to investigate the effects of stroke on the lives and feelings of patients. Accordingly, PRO assessment is a prerequisite for evidence-based and client-centered stroke rehabilitation.
To be informative, PRO assessments usually cover a few aspects of life affected by stroke, such as communication, activities of daily living, and social participation (Duncan et al., 2003; Hsueh et al., 2011). These multidimensional scores help users understand patients’ perceptions of recovery in different aspects of life. However, a major drawback of these multidimensional scores is that they cannot provide an outcome index of patients’ overall recovery (Kingsley & Patel, 2017). Although patients may achieve high levels of satisfaction in most aspects of life during the recovery process, doing so does not guarantee high ratings for their overall recovery scores (Kingsley & Patel, 2017). To demonstrate the overall recovery perceived by patients, a single score directly rated by patients is a simple and faster way to understand the outcome of stroke rehabilitation.
The overall recovery scores can be obtained by two methods. The first method is calculating the difference in two total scores summed from multiple domain scores in a particular outcome measure. For instance, the change in the sum of the 12 domain scores of the Stroke-Specific Quality of Life scale is an index that represents the overall recovery (Hsueh et al., 2011). The second method utilizes a single score, usually rated on a visual analog scale, to represent the level of recovery directly. For example, the Stroke Impact Scale 3.0 (SIS 3.0) contains an item to assess patients’ self-perceived recovery from stroke in terms of a percentage between 0% (no recovery) and 100% (full recovery; Duncan et al., 2003). Both methods have been widely used in previous studies (Cella et al., 2010; Hsueh et al., 2011).
Among these methods, the single score is more straightforward, because it directly reflects the patients’ perceptions of the recovery. Conversely, although the summation weights of each domain for the summed scores can be purposefully determined to strengthen some important domain scores, whether the summation weights can validly represent patients’ perceptions of functions is unknown. Moreover, patients may have unique perceptions of the levels of overall recovery, which should not be affected by the items and domains included in a measure. Accordingly, the single score seems to be a better index of overall recovery from stroke.
However, the clinical implications of the single score may be restricted, because the patient-reported variables that are associated with single overall recovery scores are unclear. Such information would help to reveal how patients determine their overall recovery (research value) and could be used to guide intervention plans (clinical value). Although some patients may report explanations or comments associated with their ratings, whether these considerations are affected by patients’ clinical characteristics, such as stroke severity, remains unknown. Given these uncertainties, it is difficult to fulfill either clinical or research purposes with a single overall recovery score.
The purpose of this study was twofold: to investigate which patient-reported variables contribute to the explanations of the self-reported overall recovery score in stroke patients, and to examine whether the contributing factors vary by stroke severity. Given that patients with various levels of severity may have different recovery experiences, their priorities in determining their levels of overall recovery are likely to be different.
Method
Participants
This was a secondary analysis of data extracted from the Field Administration of Stroke Therapy–Magnesium study (Saver et al., 2014). Patients were recruited between January 2005 and December 2012 if they were suspected of having stroke, were between ages 45 and 95 yr, had symptoms related to stroke for over 15 min, and had started treatment within 2 hr after the presence of the symptoms (Saver et al., 2014). The SIS 3.0 (Duncan et al., 2003) and the National Institutes of Health Stroke Scale (NIHSS; Brott et al., 1989) were completed 90 days after the presence of stroke symptoms (Saver et al., 2014). Because this study involved secondary analysis of deidentified data, ethical approval for this study was not required.
Procedure
We extracted data only for complete responses to the SIS 3.0 items reported by patients. We conducted subgroup analyses to examine the factors that influence the overall recovery scores in patients with various degrees of stroke severity. We included the data completed by the patients only, because their concerns may differ from those of their caregivers.
Measures
The SIS 3.0 is a self-report questionnaire that evaluates disability and health-related quality of life after stroke. It consists of 59 items that assess eight domains of life affected by stroke: strength, hand function, mobility, activities of daily living, emotion, memory, communication, and social participation (Duncan et al., 2003). Each item is rated using a 5-point Likert-type scale (on which 1 indicates an inability to complete the task and 5 indicates no difficulty experienced at all) to indicate the level of agreement and/or frequency of encountering difficulty in performing daily activities (Duncan et al., 2003). The item scores in each domain can be summed up as the domain score to represent the health-related quality of life in a specific aspect of life (Duncan et al., 2003). The SIS 3.0 also contains one item that assesses patients’ perception of overall recovery from stroke using a visual analog scale (Duncan et al., 2003). The SIS 3.0 has adequate reliability and validity in persons with stroke (Chou et al., 2015).
The NIHSS was used to assess the severity of stroke. It consists of 11 items that assess key responses for patients with stroke (Brott et al., 1989). The NIHSS’s total score ranges from 0 to 42; a higher score indicates greater severity (Brott et al., 1989). Briefly, scores less than 4 indicate minor severity; scores of 4 to 6 indicate mild severity, scores of 7 to 15 indicate moderate severity, and scores over 16 indicate the greatest severity (Chou et al., 2015). In general, the NIHSS shows adequate reliability and validity in persons with stroke (Dewey et al., 1999; Goldstein & Samsa, 1997; Zandieh et al., 2012).
Data Analysis
We used regression analysis with a forward selection method to identify variables associated with the overall recovery scores. The independent variables were the 59 individual SIS 3.0 items. The dependent variable was the one extra question on stroke recovery on the visual analog scale of 0% to 100%. We used the NIHSS scores in conducting the subgroup analysis. These variables were obtained 90 days after the onset of the stroke. The variance inflation factor (VIF) was used to detect collinearity, a violation of the assumption of regression analysis caused by the inclusion of highly correlated independent variables, which may lead to invalid regression models. A VIF value less than 10 was considered acceptable, because the findings would not be substantially affected (Alexander, 1994). A coefficient of determination (R 2), an index representing the percentage of variances that the selected variables can explain, was used to evaluate the effectiveness of the regression models. The variables were included on the basis of p values, using a significance threshold of .05.
Results
Of the initial data from 1,700 patients, those from 950 patients with complete responses to the SIS 3.0 questionnaire were included in the analysis. The characteristics of the participants are shown in Table 1. In general, most patients were male (61.1%) and White (77.1%), with an average age of about 70 yr. The majority of strokes were acute cerebral ischemia (79.8%). The mean NIHSS scores were 7.8 (SD = 7.2) at baseline and 1.9 (SD = 3.0) 90 days after stroke onset. For the overall recovery score, M = 74.4 (SD = 25.8). Patients with more severe stroke generally exhibited smaller overall recovery scores: Overall mean recovery scores for patients with minor, mild, and moderate stroke were 81.1 (SD = 20.8), 51.1 (SD = 22.9), and 37.5 (SD = 25.1), respectively.
Participant Characteristics (N = 950)
Note. NIHSS = National Institutes of Health Stroke Scale.
The results of the subgroup analysis are shown in Table 2. Only a few patients had severe stroke (n = 4), so they were excluded to avoid bias. The analyses were conducted in three subgroups divided by severity: minor (n = 779), mild (n = 86), and moderate (n = 81). The models explained about 60% of the variances, except for those of patients with moderate stroke (16%). As the severity of stroke increased, the number of variables associated with the overall recovery decreased. Most of the selected variables were related to social participation, followed by activities of daily living and the remaining domains of the SIS 3.0, except the memory and thinking domain. The actual variables that were selected depended on stroke severity (Table 2). None of these variables had VIF values greater than 3.0.
Variables Associated With the Overall Recovery Score According to Severity Subgroups
Note. National Institutes of Health Stroke Scale scores less than 4 were considered minor, scores between 4 and 6 were considered mild, scores between 7 and 15 were considered moderate, and scores higher than 16 were considered severe. ADL = activity of daily living; SIS 3.0 = Stroke Impact Scale 3.0; VIF = variance inflation factor.
Discussion
For most subsamples, R 2 > .50, except for those of patients with moderate stroke. Given that such R 2s are relatively large for clinical studies (Feys et al., 2000; Tashiro et al., 2021), the findings suggest that patients with subacute stroke have similar considerations in evaluating their overall recovery scores. Moreover, all selected variables showed acceptable VIF values, suggesting that the models were free from collinearity. Furthermore, the selected variables of the subsamples were similar to those of all samples, which supports that the selection of variables are robust and unbiased. Accordingly, the established regression model could be a valid and useful tool for revealing how patients perceive their overall recovery.
Several issues can be derived from the findings. First, the ability to perform self-care activities, including controlling the bowels and dressing the torso, were still a concern of patients with minor and mild stroke. The findings align with those of a previous study (Kristensen et al., 2014), indicating that patients may still perceive difficulty in performing self-care activities even if the severity of stroke is mild. Note that these variables were patient reported, which may be inconsistent with their actual abilities and real-life performances (Lee et al., 2014). Thus, although patients may have been able to complete these activities by themselves and have done so in daily life, they still perceived inconvenience in these activities. Because self-care activities are signs of independent and normal life (Jones et al., 2008), inconvenience in these activities can reasonably lead to lower ratings of overall recovery. Therefore, interventions targeting self-care activities can be considered, such as strategies to reduce the inconveniences.
Second, the identified variables were inconsistent across the subgroups divided by stroke severity. These findings support the concept that interventions should be personalized depending on patients’ characteristics, including the severity of stroke (Richards & Cramer, 2023). Moreover, the findings may help facilitate the feasibility of personalized interventions by revealing activities that are prioritized by patients with specific severity of stroke. For example, patients with minor stroke (with NIHSS scores <4) prioritized specific functional abilities, such as participation in a conversation and helping others. Patients with mild stroke (with NIHSS scores of 4–6) tended to have more general concerns, such as the roles of a family member and a friend. Patients with moderate stroke (with NIHSS scores of 7–15) prioritized the ability to help others and to enjoy things as much as ever. The identified priorities may be considered while generating personalized interventions to optimize the sense of overall recovery from stroke.
Third, the contents of most of the selected variables were related to social participation regardless of the patient’s severity, although the variables selected for each subgroup were different. For example, patients with minor stroke emphasized the ability to help others and the ability to work. On the other hand, patients with mild stroke cared more about their social roles and ability to engage in conservations. This finding is consistent with previous studies reporting that social participation is crucial for patients with stroke (Lynch et al., 2008). Therefore, clinicians may consider providing interventions that target social participation at the beginning of interventions rather than treating social participation as a secondary target.
Fourth, the R 2 of the model in the moderate subgroup (.16) was substantially lower than those of the mild severity (.51) and minor severity (.58) subgroups. The low explained variation (about 10%) was also found in a previous study of patients with moderate stroke (Johnston et al., 1999). One explanation is that the perception of overall recovery was individualized, with limited similarity across patients, so the explained variance was small. However, because the explained 16% of variance was significantly different from zero, the validity of this explanation is questionable. Another explanation is that the key factors associated with the overall recovery scores in patients with moderate stroke were not covered by the SIS 3.0 (Lin et al., 2010). Therefore, future studies can include more variables that are related to self-perceived overall recovery in patients with moderate and even severe stroke. On the basis of the present findings, the ability to “enjoy things as ever” and the ability to “help others” seem to affect patients’ perceptions of overall recovery and may be integrated into routine interventions in patients with moderate stroke.
Some patient-reported variables showed negative regression coefficients in the models, such as the ability to enjoy things as ever (b = −4.8) for patients with moderate stroke, remain standing (b = −3.3) for those with mild stroke, and control the bowels (b = −2.8) for those with minor stroke. These findings indicate that when the influences of other variables are controlled for (Nagelkerke, 1991), patients’ overall recovery scores decrease as their satisfaction with these variables increases. The negative correlations may be attributed to the distinct nature of these patient-reported variables, such as the formulation of reversed items, or to patients’ rationalization processes poststroke (Lee et al., 2021; Suárez-Álvarez et al., 2018). Given that full recovery 90 days poststroke may not be feasible, patients may lower their expectations, enabling them to derive satisfaction from everyday activities despite not fully recovering (Groeneveld et al., 2019). Conversely, patients with less severe strokes may maintain higher recovery aspirations, leading to unsatisfaction with the ability to perform basic tasks, such as standing or controlling one’s bowels. These dynamics suggest the importance of incorporating an evaluation of self-efficacy and recovery expectations in future studies to better understand the underlying causes of negative coefficients observed.
It is interesting that no variables were selected from the memory and thinking domain. Even in the results of subgroup analyses, the memory and thinking items were excluded. These findings may have three explanations. First, memory and thinking are not the major concern of patients with minor to moderate stroke. However, approximately 60% of patients have cognitive impairment in at least one domain (Nakling et al., 2017), suggesting that this explanation is unreliable. Second, key variables of the memory and thinking domain are not included in the SIS 3.0. Third, adequate memory and thinking abilities are the prerequisite of patient-reported data. Thus, more items and patients with various levels of cognitive function may be included to examine the role of memory and thinking in self-perceived overall recovery.
Two advantages may reinforce the robustness of this study. First, the sample size was large, so the findings can be expected to be reliable. Second, the data were collected at just 90 days after stroke. Therefore, the variations caused by inconsistent timeframes have been ruled out. In addition, the variables associated with overall recovery scores overlapped between the models established for all participants and each subsample, so the findings seem to be stable. Accordingly, the results may be reliable and valid for investigating the variables associated with overall recovery from stroke.
Study Limitations
Some limitations should be noted. First, few participants with severe stroke (NIHSS scores >16) were included in this study, which may be due to patients having difficulty in completing the self-reported questionnaires. Therefore, the findings may not be generalizable to patients with severe stroke. Second, the adopted data were cross-sectional. Thus, the findings do not provide direct evidence of the effectiveness of interventions targeting these identified variables on overall recovery scores. Third, most participants were White male patients with ischemic stroke. Thus, the findings may not apply to all patients of other genders, races, and types of strokes. The influences of these characteristics on the overall recovery can be further investigated. Fourth, the SIS 3.0 items were rated using a 5-point Likert-type scale, so the scores may have had floor and/or ceiling effects that could affect the results of regression analyses (Hawe et al., 2019). In future studies, cross-validation with items and/or variables rated on a visual analog scale to achieve continuous scores free of ceiling and floor effects is warranted. Moreover, variables that are associated with future overall recovery may be future targets of the investigation.
Implications for Occupational Therapy Practice
This study has the following implications for occupational therapy practice: This study empirically supports occupational therapy practitioners’ interventions that target patients’ social participation and self-care activities. The identified variables across the severity groups may help occupational practitioners plan intervention programs and optimize patients’ self-perceived levels of overall recovery from stroke.
Conclusion
The patient-reported variables associated with the overall recovery scores differed, depending on the stroke severity. Most variables were related to social participation and self-care activities, particularly for the abilities to help others, control the bowels, and dress the torso. On the basis of these findings, clinicians might set treatment goals and design interventions by considering the variables for the severity subgroups to optimize patients’ self-perceived overall recovery.
Footnotes
Acknowledgments
Hsin-Yu Chiang and Ching-Lin Hsieh contributed equally to this work and serve as corresponding authors. We have no conflicts of interest to disclose. We all agree with the stated authorship of and contributions to this article. This research was supported by National Science and Technology Council (109-2314-B-002-110-MY3).