Stroke Impact Scale Version 2: Validation of the French Version
- Sylvia Caël,
- Pierre Decavel,
- Christine Binquet,
- Charles Benaim,
- Marc Puyraveau,
- Morgane Chotard,
- Thierry Moulin,
- Bernard Parratte,
- Yannick Bejot and
- Mariette Mercier
- S. Caël, MD, Department of Biostatistics–EA 3181, Faculty of Medicine, University of Franche-Comté, 19 rue Ambroise Paré, 25000 Besançon Cedex, France.
- P. Decavel, MD, Department of Rehabilitation, University Hospital of Besançon, Besançon, France.
- C. Binquet, MD, PhD, University Hospital of Dijon, Clinical Investigation Center, Clinical Epidemiology/Clinical Trials, Dijon, France; INSERM, CIE1, Dijon, France.
- C. Benaim, PhD, Department of Rehabilitation, University Hospital of Dijon.
- M. Puyraveau, MSc, University Hospital of Besançon, Clinical Methodology Center.
- M. Chotard, University Hospital of Besançon.
- T. Moulin, MD, PhD, Department of Neurology, University Hospital of Besançon.
- B. Parratte, MD, PhD, Department of Rehabilitation, University Hospital of Besançon.
- Y. Bejot, MD, Department of Neurology, University Hospital of Dijon.
- M. Mercier, PhD, Laboratoire de Biostatistique EA 3181 IFR 133, University Hospital of Besançon and University of Franche-Comté.
- Address all correspondence to Dr Caël at: sylviacael{at}yahoo.fr.
Abstract
Background Almost 1 person in 1,000 experiences a stroke annually in France. Health-related quality-of-life (HRQoL) measurement with specific questionnaires is useful to study the consequences of stroke on patients' daily lives.
Objective The purpose of this study was to validate the French version of the Stroke Impact Scale (SIS) questionnaire, as no disease-specific questionnaire was validated in French heretofore.
Methods Two hundred eighty-eight patients with stroke were classified into 2 groups (158 acute, 130 chronic). Rate of item completion, test-retest reliability, sensitivity to change, and construct validity, convergent validity, and discriminant validity of the questionnaire were assessed. Acute group patients were recruited during the first month poststroke and followed for 3 months. Chronic group patients (stroke dating from 1 year) were recruited from outpatient consultations. The first 100 patients in the chronic group were called back 15 days after inclusion for test-retest measurements. The Barthel Index, Hospital Anxiety and Depression Scale (HADS), and Duke Health Profile questionnaires were administered.
Results The French version of the SIS was well accepted by all participants. It had good reproducibility. Cronbach alpha was 89% for all scales. A ceiling effect was noted in the majority of scales. Physical domains were significantly correlated to other measures of physical capacity (Barthel Index and Duke Health Profile; Spearman coefficients were between .5 and .73), and the emotional and social domains were significantly correlated to almost all domains of the Duke Health Profile. Psychometric properties were similar to those of the US version. Responsiveness was good for physical and emotional domains.
Limitations Validation was conducted only on people who were able to answer the questionnaire.
Conclusions The French SIS version presents good psychometric properties, except for ceiling effect. This is the first stroke-specific questionnaire evaluating perceived health after stroke to be validated in France and could be useful for further investigations of HRQoL after stroke.
As in all industrialized countries, France is confronted with an increasing frequency of cerebrovascular events. Indeed, stroke is the fifth most prevalent cause of moderate to severe physical disability in people under 59 years of age, and the third most prevalent cause above the age of 60 years in high income countries.1 The incidence of stroke is almost 1 per 1,000 people per year in France2 and can lead to severe consequences in patients' everyday life. Given that the life expectancy of patients after stroke has increased,2 stroke can represent a long-lasting burden for both the patient's family and society in general. Health-related quality-of-life (HRQoL) questionnaires are useful tools to evaluate a patient's state of health, in its totality, from the patient's point of view. Unfortunately, no specific HRQoL questionnaire for stroke has undergone complete validation in a French-language version up to now (although a translation into French was available), yet such a tool is needed to prove the benefit yielded from pharmacological treatment and rehabilitation.
The purpose of this study was to validate the French version of the Stroke Impact Scale (SIS) version 2, which is an HRQoL self-report questionnaire developed by the University of Kansas.3 This scale was translated into French using appropriate methods (translation and back-translation) by Mapi Research Trust, a nonprofit organization in France, but its psychometric properties remained to be validated. In the context of transcultural adaptation, it is recommended that the psychometric properties of translated tools be verified.4 For classical validation and to evaluate both reliability and sensitivity to change, a sample comprising both stable and unstable patients was required. To this end, 2 groups were considered, namely “acute” and “chronic.” We hypothesized that patients in the acute group would experience a change in their symptoms over the first 3 months after the initial event, whereas those in the chronic group were expected to remain stable. This study is the first phase of a more ambitious project to create a French cohort (called Hemiqual) of patients who experienced stroke and to describe the development of their quality of life over 18 months and factors associated with quality of life in this population. Moreover, data on sensitivity to change were lacking in the US validation. Therefore, we also aimed to estimate sensitivity to change of the SIS questionnaire.
Method
Participants
This validation study was conducted between March 2010 and May 2011 in 2 regions in eastern France: Burgundy and Franche-Comté (combined population of 2.8 million inhabitants) (eFigure).
Consecutive patients were recruited in the neurology and rehabilitation departments of the university hospitals of Dijon and Besançon, which are the main referral centers for stroke.
Inclusion and Exclusion Criteria
Stroke was defined as a neurologic deficit of sudden onset, whose vascular etiology was confirmed by computed tomography or magnetic resonance imaging. Both patients with ischemic and hemorrhagic strokes were included, as well as those with first and recurrent strokes. Patients with acute stroke were recruited in the first month after stroke. Patients in the chronic group whose stroke dated from more than 1 year previously were recruited among outpatient consultations in neurology and rehabilitation. Exclusion criteria were: age <18 years, previous invalidity (data on previous invalidity status were obtained by consulting each patient's medical file), and inability to complete the assisted questionnaire (ie, dementia or Boston Diagnostic Aphasia Examination [BDAE] score <35). Employment was classified as currently working versus out of work (unemployed or invalid or retired or voluntarily not working). All participants gave informed consent.
Assessments
The Stroke Impact Scale (SIS) version 2 is a specific HRQoL questionnaire comprising 8 domains (63 items plus a recovery scale): strength (4 items), memory (8 items), emotion (9 items), communication (7 items), activities of daily living/instrumental activities of daily living (ADL/IADL; 12 items), mobility (10 items), hand function (5 items), and social participation (8 items). The physical function score is a sum of 4 scales (strength, ADL, mobility, and hand function). Scores are normalized between 0 and 100.3,6,7 Version 2 was preferred to the short version (SIS-16) in our study because the short version is more representative of physical capacities than HRQoL.6
We chose to evaluate version 2, and not version 3, in order to be able to compare psychometric properties with the US version 2. Indeed, at the time our study was designed, no report of the psychometric properties of the English-language SIS version 3.0 was yet available. The US validation reported the followed properties3: (1) construct validity, showing 5 eigenvectors in factor analysis: physical, emotional, communication, memory, and social participation; (2) ceiling effects were observed, with intraclass coerrelation coefficients (ICCs) ranging from .57 (emotional domain) to .92 (physical domain); (3) Cronbach alpha coefficients were between .83 and .90; and (4) convergent validity, showing high correlations with other physical and emotional scales but lower correlations for communication and memory. The social participation domain showed a Spearman coefficient of .7 with the social domain of the 36-Item Short-Form Health Survey (SF-36). Sensitivity to change was not evaluated, only a significant time effect was described, and no comparison between patient groups was reported regarding the change in their health state. Mapi Research Trust (Lyon, France) performed the translation and back-translation of the SIS questionnaire. All relevant permissions for validation were obtained. The SIS questionnaire and permission to use it can be obtained from Mapi Research Trust, Lyon, France, on behalf of the copyright holders, the University of Kansas Medical Center (Kansas City, Kansas).
Because patients with stroke are often disabled, they were assisted for the questionnaire administration by a trained assistant who read out the questionnaires, ticked the patient's chosen answer, and could explain any unclear points according to the administration guidelines. In the acute group, questionnaires were administered during the first month poststroke and a second time 3 months later. Patients who failed to attend the 3-month visit were contacted by telephone for administration of the questionnaire. In the chronic group, questionnaires were given at inclusion. The first 100 patients included in the chronic group were contacted by telephone 2 weeks later, and the questionnaire was administered by telephone.
We recorded sociodemographic characteristics and clinical data (National Institutes of Health Stroke Scale [NIHSS] scores, stroke type and location). The NIHSS is a standardized instrument to measure the clinical level of impairment caused by stroke.8 The higher the score, the more severe the deficit. In our study, we chose a cutoff value of 6, as in a previous study by Adams et al.9 People with an initial NIHSS score of <6 were considered as having had a minor stroke (versus NIHSS scores of ≥6 defining moderate and severe stroke); people who were in a comatose state (initial NIHSS score not available) were grouped with those with moderate or severe stroke.
In addition, all participants answered the following questionnaires: Hospital Anxiety and Depression Scale (HADS), Barthel Index (BI), Duke Health Profile, and Color Analog Scale (CAS). The HADS indicates the level of anxiety and depression symptoms. Both anxiety and depression subscales were scored between 0 and 21, with a higher score indicating severe anxiety or depression symptoms.10 The BI is a scale measuring independence. Possible scores range from 0 to 100, and patients who score over 95 are considered to be independent.11 The Duke Health Profile is an HRQoL questionnaire comprising 8 domains and has been validated in the French language.12 The CAS is a vertical red analog pain scale, which is adapted to each side of the stroke lesion.13 For all instruments, scores were calculated according to the developers' guidelines. Missing data were completed according to the developers' guidelines for the SIS and Duke Health Profile, using means substitution for the HADS, and were not completed for the CAS.
Data Analysis
Rates of missing items and complete questionnaires were calculated. Participants were asked a range of questions relating to their perception of the questionnaires (eg, too long, questions embarrassing or hard to understand). Concerning construct validity, confirmatory principal components analysis was used to verify the item's clustering in domains. Cronbach alpha, floor and ceiling effects, construct validity, convergent validity, and discriminant validity were evaluated using all questionnaires at 3 months in the acute group, plus the questionnaires completed at inclusion in the chronic group. Internal consistency was evaluated using Cronbach alpha coefficient and floor and ceiling effects. The reproducibility was assessed by the test-retest method using ICCs among participants in the chronic group, who were contacted by telephone 2 weeks later to complete the questionnaire a second time by telephone.14,15 Convergent validity was evaluated using Spearman rank correlation coefficients between the SIS and other questionnaires and between SIS domains. Results are presented in a multitrait-multimethod matrix. The Kruskall-Wallis test between groups was used to evaluate discriminant validity. Sensitivity to change was analyzed among participants of the acute group using score variations and the standardized response mean (SRM).16 Participants whose Duke Health Profile physical health score remained stable or decreased were included in the “worsening physical health” group, and those whose Duke Health Profile physical health score increased were included in the “improved physical health” group. Similarly, participants whose HADS score increased or remained stable were included in the “worsened mood” group, and those whose HADS score decreased were included in the “improved mood” group. Sensitivity to change was evaluated using the SRM (mean of score difference divided by the standard deviation of the score difference).
Sample Size Calculation
To perform a factor analysis, about 300 patients were needed.17 In order to show a 10-point difference with an alpha level of .05 and 80% power,18 75 participants per group were needed to evaluate sensitivity to change. Thus, the number of participants required was 160 in the acute group and 140 in the chronic group.
All analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, North Carolina). All tests were 2-sided, and a P value of <.05 was considered significant.
Role of the Funding Source
This work was supported by an interregional grant from the French National Programme for Clinical Research (Programme Hospitalier de Recherche Clinique [PHRC]).
Results
Study Participants
Overall, 311 participants met the inclusion criteria and were eligible for inclusion, of whom 288 (93%) agreed to participate (158 in the acute group and 130 in the chronic group). The sociodemographic and clinical characteristics of the study participants are presented in Table 1. For patients with minor stroke, the mean NIHSS score was 2.3 (SD=1.7, range=0–5); there was no difference between the US and French versions. For patients with moderate to severe stroke, the mean NIHSS score was 11.6 (SD=5.1) versus 6.6 (SD=4.5) in the US version (P<.0001). Among the 23 patients who declined to participate (16 [9%] in the acute group and 7 [5%] in the chronic group), there were significantly more men (75% male), of significantly older age (mean age=75 years), but the initial NIHSS score did not differ from that of the patients who were included. In the chronic group, 27 patients failed to answer at 15 days; however, age, sex, and NIHSS score were not different from patients in the chronic group who responded to the second questionnaire. In the acute group, at 3 months poststroke, 6 patients (3.8%) were lost to follow-up and 3 had deteriorated physically such that they were no longer able to participate.
Sociodemographic and Clinical Characteristics of the Study Participantsa
Rate of Item Completion
The rate of item completion was 98.4% in the chronic group and 94.6% and 97.3% in the acute group at inclusion and at 3 months, respectively. Fifty-six percent of the questionnaires were fully completed in the chronic group at inclusion and 64% at 15 days. In the acute group, 29% of the questionnaires were fully completed at inclusion and 43% at 3 months. The most frequently missing items were those regarding spiritual or religious activities (28% and 18%, respectively, in the acute group at 3 months and in the chronic group), work (12% and 14%, respectively, in the acute group at 3 months and in the chronic group), and active recreation (12% and 4%, respectively, in the acute group at 3 months and in the chronic group). The work item was significantly more frequently missing when participants had no professional activity (P=.02) in the chronic group.
In terms of acceptability of the questionnaire, a quarter (67/266) of all patients responded that they would have liked to be asked about other topics (mainly sexuality [10 patients] and fatigue [9 patients]); 27% (75/273) found at least one question difficult to understand; and only 6% found at least one question embarrassing (especially in the social domain). Eighty-three percent found the length of the questionnaire to be acceptable.
Reliability
Cronbach alpha coefficients were over 89% in all domains (Tab. 2). A majority of domains suffered from a ceiling effect for patients with minor stroke. Memory, communication, and hand function presented ceiling effects for patients with moderate to severe stroke. The ICCs were all above .65, reaching .92 for physical function (Tab. 2).
Reliability of the French Version of the SIS, Version 2: Cronbach Alpha Coefficient, Floor and Ceiling Effects, and ICCsa
Construct Validity
Principal component analysis showed grouping of items into 8 domains (Appendix). Two items concerning sphincter function were poorly correlated to the ADL domain. Three emotion items, for which the Likert response scale was reversed, were grouped together. With the exception of these 2 points, all other item grouping appeared to be adequate in the 8 domains. Strength, ADL, mobility, and hand function could be clustered in the physical function domain. Table 3 shows the correlations among the scores of SIS domains. The SIS physical domains showed good correlation with other SIS physical domains, and memory and communication were more highly correlated with each other and with social participation than with the physical domains.
Multitrait-Multimethod Matrix: Spearman Correlation Coefficient Between SIS Domain Scores and HADS, BI, and Duke Health Profile Scoresa
Convergent Validity
Correlation coefficients with HADS, BI, and Duke Health Profile are presented in Table 3. There were no significant correlations between SIS domains and NIHSS or pain scales. All physical domains were correlated with the BI and with the Duke Health Profile physical domain, with a Spearman correlation coefficient over .54. Emotion was correlated (Spearman correlation coefficient=.54–.64) with HADS score and with almost all Duke Health Profile domains. Social function was correlated with the HADS depression scale, BI, and almost all domains of the Duke Health Profile (except the social domain).
Discriminant Validity
The SIS scores were statistically and clinically different (they could reach 30 points, especially in physical domains) according to BI, stroke severity (for physical and social domains), and BDAE score (except emotion) (Tab. 4). A majority of scores were different depending on whether patients had a professional activity. Strength, mobility, and hand function were different according to stroke location, with lower scores in patients with minor hemisphere lesions. No SIS domain was discriminant between groups for stroke type, sex, or age.
Discriminant Validity: Mean SIS Scores Between Patient Groupsa
Sensitivity to Change
The domains that were most sensitive to improvements in physical health were ADL, mobility, and hand function (with SRM above 0.5 and an absolute difference of 16 points), as well as the physical function domain, with an SRM of 0.9. Along with social participation, the physical domains also were most sensitive to mood change. The SRM of the emotional domain was 0.42 for patients in the improved mood group and −0.54 for patients in the worsened mood group. Results of the sensitivity to change analysis are presented in Table 5 (physical health) and Table 6 (mood).
Sensitivity to Change in Degree of Autonomy Between First Assessment and Reassessment at 3 Months in Participants in the Acute Groupa
Sensitivity to Change in Mood Between First Assessment and Reassessment at 3 Months in Participants in the Acute Groupa
Discussion
The SIS is the first stroke-specific HRQoL questionnaire to be validated in French. Our results show that the French version is reliable (except for a ceiling effect), discriminant, sensitive to change, and well correlated to other scores. Overall, the French version of the SIS showed good sensitivity to change for both autonomy and mood. A strong point of this study is the fact that the SRM and differences between groups based on patient evolution are evaluated here for the first time for the SIS. The considerable differences observed suggest that this questionnaire is capable of showing change in both physical and mental aspects of perceived health. Thus, it is possible to use the SIS to evaluate the benefit of a therapeutic intervention, such as rehabilitation or mental health (eg, via drugs or psychotherapy), that aims to improve physical health.
A limitation of the validation design is the language specificity, in that this version will naturally only be applicable in French-speaking countries. However, cultural nuances in the way French is spoken across different French-speaking countries of the world would likely be only minor, and adjustments could easily be made without having to undertake a whole new process of cross-cultural adaptation.
Recruitment was rapid, with few patients declining participation. The inclusion and exclusion criteria were as wide as possible in order to validate the SIS in as large a population as possible and to make it widely applicable. In our sample, the patients were relatively young (5 years younger, on average, than in the population used for the US validation), especially in the chronic group, but these were outpatients who had survived more than 1 year. In any case, the mixture of both acute and chronic populations does not appear to have affected the results. The initial low mean NIHSS score (<6) could be explained by the inclusion criteria, which were designed to identify people who would be able to answer the assisted questionnaire, whereas the US validation concerned minor and moderate stroke.3 The most frequently missing questions were about religion, work, and recreational activity. Possible explanations are that the participants may have had difficulty understanding some questions (27% [75/273] found at least one question difficult to understand). Alternatively, perhaps the patients did not feel affected by these items. Indeed, 28% of the French population declares that they have no religion.19 Similarly, the fact that people in the chronic group who were not working more frequently failed to answer the question about professional activity could indicate that they did not feel concerned. It is noteworthy that this difference was not significant in the acute group. However, missing items did not prevent global score calculation, because there were very few missing items per questionnaire. Item grouping into domains seemed to be acceptable and was similar to the US version (especially the clustering of 4 domains into a physical function domain). The 3 items of the emotion domain were to be answered with an inverted Likert scale, and these questions were often misunderstood by the participants and answered as if the Likert scale had been the other way round. Drawing the patient's attention to the inversion of the Likert scale could be recommended for future use.
The French version suffered a ceiling effect for patients whose initial NIHSS score was <6, more so than in the US version.3 However, stroke severity was not defined on the same scale in our study (Orpington Prognostic Scale versus NIHSS3). Initial NIHSS scores were not significantly different between the US and French versions for patients with minor stroke but were significantly higher in the French version for the moderate stroke group. However, comparison is difficult because the categorization of stroke severity was different between the US and French versions, and it is possible that the French version suffered a greater ceiling effect than the US version.
Reproducibility and Cronbach alpha were similar to those of the US version.3 Two ICCs were lower than .7, but the values were nonetheless close to, and even better than, in the US version (emotional domain: .57 for the US version versus .67 for the French version). Correlations with other scales also were generally logical and quite strong, except for the social domain, where the correlation with the Duke Health Profile social domain was low (.43). The US social domain correlated well with the SF-36 social domain (.7). Communication had a low correlation, except for the Duke Health Profile global health domain. Conversely, the only other available source of information on communication, namely the BDAE, is a qualitative scale; therefore, no correlation could be tested between the SIS domains and BDAE because only 13% of the patients had a BDAE score of <5. Nonetheless, the discriminant validity of communication was good. Other SIS domains showed good to excellent discriminant validity, especially as regards autonomy level. Other studies have confirmed that HRQoL (evaluated using other questionnaires) after stroke is influenced by stroke severity20,21 and patient autonomy.22–24 There was no significant difference in SIS scores according to age and sex. Some studies have shown that women have a lower HRQoL score after stroke,20,21,25,26 whereas one study described lower HRQoL scores in men.27 Other studies have demonstrated lower HRQoL scores with older age.20,21 With the SIS, we observed that patients with lesions of the minor hemisphere had lower scores. This finding is at odds with a previous study by Nichols-Larsen et al,20 who observed lower scores in cases of dominant side lesions.
For clinicians, quality-of-life studies represent an opportunity to take an interest not just in the patient's illness but also in his or her life in general. One weakness of the SIS questionnaire is that it does not ask about fatigue and sexuality. Clinicians could broach these subjects with patients during follow-up visits, and further studies or questionnaires could consider including these 2 points.
In conclusion, the French version of the SIS demonstrated satisfactory psychometric properties and will be useful for future HRQoL studies in French-language populations. The methodology of this validation of sensitivity to change should be replicated in further studies to ensure that the questionnaire can detect differences. The minimal clinically important difference remains to be validated in French.
Appendix.
Principal Component Analysis of the Stroke Impact Scale Version 2, French Version
Commentary: Memory and communication were positively correlated with the same principal component (number 2), but correlations with component 7 were the opposite of each other.
For items 6, 8, and 9 of the emotion domain, the correlation coefficient inverted between principal components 7 and 8. In these particular items, the Likert scale was reversed, and many respondents did not notice or understand, thereby answering as if the Likert scale was the other way around.
For example:
Item 5: Blame yourself for mistakes that you made?
Item 6: Enjoy things as much as ever?
Item 7: Feel quite nervous?
Strength and hand function were correlated to principal component 4.
The activities of daily living (ADL) domain is separated into 2 parts, namely ADLs (items 1–7) and instrumental activities of daily living (IADLs) (item 8–12). Items 6 and 7 relate to sphincter function and had the highest correlation coefficient. ADLs were correlated with principal component 9, whereas IADLs and the mobility domain were correlated with the first principal component.
The social participation domain is represented by the fifth component.
No comparisons can be made with the US version; the principal component analysis of the US version has not been published.
Footnotes
Dr Caël, Dr Binquet, Dr Benaim, and Dr Mercier provided concept/idea/research design and writing. Dr Caël, Dr Decavel, Miss Chotard, Dr Moulin, and Dr Bejot provided data collection. Dr Caël, Dr Benaim, Mr Puyraveau, and Dr Mercier provided data analysis. Dr Caël, Dr Benaim, and Dr Mercier provided project management. Dr Caël and Dr Mercier provided fund procurement. Dr Benaim, Miss Chotard, Dr Moulin, Dr Parratte, and Dr Bejot provided study participants. Dr Benaim, Dr Moulin, Dr Parratte, and Dr Bejot provided facilities/equipment. Dr Binquet and Dr Mercier provided institutional liaisons. Miss Chotard provided administrative support. Dr Decavel, Dr Binquet, Dr Benaim, Mr Puyraveau, and Dr Bejot provided consultation (including review of manuscript before submission).
The authors thank Emilie Galizzi and Sandrine Vinault for their help with computer programming, Véronique Cote and Dr Didier Chavot for their kindness and help with inclusions, and Fiona Ecarnot (EA3920, University Hospital of Besançon, France) for translation and editorial assistance.
The Ethics Committee of the University Hospital of Besançon and the national committee for the protection of privacy and personal data (Commission Nationale de I'Informatique et des Libertés [CNIL]) approved the study.
This work was supported by an interregional grant from the French National Programme for Clinical Research (Programme Hospitalier de Recherche Clinique [PHRC]).
- Received February 4, 2013.
- Accepted November 23, 2014.
- © 2015 American Physical Therapy Association