French validation of the Stroke Specific Quality of Life Scale (SS-QoL)

Abstract

OBJECTIVE:

To adapt the SS-QoL into French and test its psychometric properties.

METHODS:

Seventy-seven patients from a population-based registry were enrolled 3 months after their stroke. SS-QoL, NIHSS score, Barthel index, HAD, FSS, SF-36 scales, and MMSE were administered at enrolment. SS-QoL was re-administered at 15 days and 2 months. Internal consistency was assessed by Cronbach’s α coefficients, factorial validity by an exploratory factor analysis and external validity by Mann-Whitney test and Spearman’s correlations (ρ), comparing SS-QoL scores with those obtained from established scales. Reliability was assessed by intra-class correlation coefficients (ICC) and responsiveness by standardized effect sizes (ES).

RESULTS:

Test-retest and inter-observer reliabilities were excellent (ICC> 0.88). Internal consistency was acceptable (α= 0.65–0.91), except for the Personality domain (α= 0.58). Factor analysis individualized eight homogenous axes. SS-QoL scores were different between groups opposed by their modified Rankin score at enrolment or their overall quality of life compared with pre-stroke status (p < 0.001). Ten of the twelve domains correlated moderately (ρ> 0.35) to strongly (ρ> 0.5) with established measures. Nine domains were mildly to moderately responsive to change (ES> 0.3).

CONCLUSION:

The French version of the SS-QoL is a valid, reliable and moderately responsive instrument.

Keywords

Stroke quality of life outcome scale validation

1 Introduction

Despite major improvements in acute care and rehabilitation in recent decades, stroke remains a disabling disease (Béjot, Rouaud, Durier, Caillier, Marie, Freysz, Yeguiayan, Chantegret, Osseby, Moreau, & Giroud, 2007; Miller, Murray, Richards, Zorowitz, Bakas, Clark, & Billinger, 2010; Kunst, Amiri, & Janssen, 2011; de Peretti, Grimaud, Tuppin, Chin, & Woimant, 2012; Koton, Schneider, Rosamond, Shahar, Sang, Gottesman, & Coresh, 2014). Currently, the evaluation of post-stroke impairment in epidemiological studies, or as an outcome of clinical trials mostly relies on functional scales such as the modified Rankin score (mRS) or the Barthel Index. Although very useful, such assessments are limited by the fact that they do not entirely reflect the real impact of stroke on patients’ lives since subjective adverse consequences of stroke including those on family, social, and occupational life, mood, or fatigue may remain unexplored by these scales (Ali, Fulton, Quinn, & Brady, 2013). Efforts are needed to create a comprehensive assessment of patients’ outcomes after stroke. Quality of life (QOL), as a multidimensional and patient-centered concept (World Health Organization, 1993), appears to be a relevant endpoint to achieve this goal. Specific scales to measure QOL are therefore needed, but only a few have been developed. Only the Stroke Impact Scale (SIS) has been validated in French (Caël, Decavel, Binquet, Benaim, Puyraveau, Chotard, Moulin, Parratte, Béjot, & Mercier, 2015), but its use in clinical practice is hampered mainly because it takes a long time to complete, especially in vulnerable patients. The Stroke Specific Quality of Life Scale (SS-QoL) (Williams, Weinberger, Harris, Clark, & Biller, 1999) appears to be an interesting alternative to the SIS. It is more synthetic than the SIS, easier to use and understand, with a smaller number of items but a wider range of explored domains as Vision, Personality and Energy domains. Also, the patient approach used in the process of item development justifies the choice of the SS-QoL. Although there are other specific scales, such as the Health Related Quality of Life In Stroke Patients-40 scale (Ojo Owolabi, 2010) or the Stroke Adapted Sickness Impact Profile-30 scale (Van Straten, De Haan, Limburg, Schuling, Bossuyt, & van den Bos, 1997), the SS-QoL has been used and quoted much more widely, and cross-culturally adapted in international stroke publications.

The aim of this study was to translate the SS-QoL, while adapting it to French cultural characteristics (linguistic validation), and to test its psychometric properties in terms of validity, reliability and responsiveness in stroke patients.

2 Methods

2.1 Study population

Consecutive stroke patients were recruited from the Dijon Stroke Registry, an ongoing population-based study that collects all cases of stroke occurring within the population of the city of Dijon, France (Béjot, et al., 2007; Béjot, 2015). From 1st November 2014 to 31st October 2015, patients were enrolled three months after the index stroke during a scheduled follow-up visit at an outpatient clinic or rehabilitation center. This period was chosen because in clinical trials on acute stroke treatment patients’ outcomes are usually evaluated at 3 months. Informed consent to participate in the study was collected from all of the patients.

Patients were enrolled if they were 18 years old or older and had suffered a first-ever symptomatic stroke (ischemic stroke or spontaneous intracerebral hemorrhage) between 1st August 2014 and 31st July 2015. Exclusion criteria were a history of previous symptomatic stroke, severe visual or hearing disability, severe aphasia, inability to read or speak French, dementia, psychiatric disorders, addiction to alcohol or drugs, bedridden state and other serious conditions impairing the ability to complete the questionnaire.

2.2 Linguistic validation of the French version of the SS-QOL

The SS-QoL is a disease-specific QOL measurement tool that includes 49 items across 12 domains (Williams, et al., 1999): mobility, energy, upper extremity function, work/productivity, mood, self-care, social roles, family roles, vision, language, thinking and personality. Each item is scored from 1 to 5 according to several possible answers (1 = lowest quality of life, 5 = maximum quality of life). The total score ranging from 1 to 5 is calculated from mean scores for each domain.

Permission to develop a French version of SS-QoL was obtained from the original corresponding author (Williams, et al., 1999). The translation and cross-cultural adaptation were performed according to standard recommendations (Epstein, Santo, & Guillemin, 2015). First, the SS-QoL was translated from English into French by two independent bilingual translators whose first language is French (DC and JPC). Second, a debriefing meeting including both translators (DC and JPC) and investigators (NL and YB) was held so as to compare the two translations, to discuss differences and propose equivalencies. At the end of this stage, a new French version was obtained by consensus. Third, a back translation of the last version was done by a bilingual translator, whose first language is English and who had not participated in the first translation (PB). Last, a review of the back translation and analysis of translations was organized during a meeting that included all of the members. Idiomatic, semantic or conceptual equivalencies were discussed and possible amendments were consensually adopted to reach the final questionnaire.

To assess content validity, the scale was presented to stroke neurologists of the University Hospital of Dijon so as to evaluate the relevance of questions and response modalities, the understanding of items by the target population, the non-ambiguity of the vocabulary and the overall consistency of the scale compared with the original one. A pre-test was conducted by submitting the final version of the questionnaire to a sample of 10 stroke patients. Understanding of the questionnaire was assessed by asking patients to reformulate items and possible difficulties or remarks were noted.

2.3 Data collected

At registration in the Dijon Stroke Registry, the following variables are collected: demographics, vascular risk factors, pre-stroke dependency according to the modified Rankin Scale (mRS) score, type of stroke (ischemic versus hemorrhagic), initial clinical severity according to the National Institute of Health Stroke Scale (NIHSS) score, acute treatment with intravenous thrombolysis.

In the present study, patients were interviewed face-to-face at enrolment during the three-month follow-up visit, and the following clinical scores were determined: NIHSS, mRS, Barthel index, Mini Mental State Evaluation (MMSE), Fatigue severity scale (FSS), Hospital Anxiety Depression scale (HAD), General Health Survey Short Form 36 (SF-36) score, SS-QoL.

To validate the reliability and the responsiveness of the SS-QoL, a sample of 60 patients (30 patients for the test-retest and 30 patients for the inter-observer reliability) were contacted again for a telephone interview 2 weeks after inclusion and again 2 months after enrolment (responsiveness).

2.4 Statistical analyses

All psychometric parameters were assessed in accordance with the internationally recommended taxonomy (Mokkink, et al., 2010). The analyses were performed with SAS^® software (version 9.4, SAS Institute, Inc., Cary, NC, USA) and STATA^® 10.0. The level of statistical significance was set at p < 0.05. Internal consistency was determined using Cronbach’s alpha coefficient (α), which was calculated for each dimension. An α≥0.7 was considered good (Nunally, 1978; Lance, Butts, & Michels, 2006). Structural validity was assessed by an exploratory factorial analysis with a varimax rotation on the first 12 principal components. We analyzed the factorial axes highlighted by this method. Correlation coefficients (or standardized load) of the items for each factorial axis determined the scale structure. The expected result was convergence of this factorial structure to the expected content of the scale. A standardized load≥0.5 in the correct domain was considered satisfactory. For external validity, convergent validity was determined using the Spearman correlation test (Streiner, & Norman, 1995; Polit, 1996; Fayers, & Machin, 2000; Lance, et al., 2006), and by comparing SS-QoL scores with those obtained from established scales. Spearman coefficients ρ< 0.35, 0.35–0.5 and >0.5 were considered low, moderate and high, respectively (Devilliers, Amoura, Besancenot, Bonnotte, Pasquali, Wahl, Maurier, Kaminsky, Pennaforte, Magy-Bertrand, Arnaud, Binquet, & Guillemin, 2012). Divergent validity was based on known-group comparisons. We first studied SS-QoL scores from groups defined by their mRS at enrolment (mRS = 0-1 vs mRS > 1). We then repeated the analysis, defining groups according to their self-reported overall quality of life compared with their pre-stroke status (worse versus not worse, using a Likert scale). In both analyses, we hypothesized that SS-QoL scores would be significantly different between groups. The nonparametric Mann-Whitney test was used (Polit, 1996). Test-retest and inter-observer reliabilities were assessed by intra-class correlation coefficients (ICC), by comparing SS-QoL scores at enrolment with those determined 2 weeks later. An ICC greater than 0.70 indicated very good reliability (Fayers, & Machin, 2000). Responsiveness was determined by the effect size (ES) and the standardized response mean (SRM). ES was calculated in patients who reported a change in their quality of life between the enrolment visit and their pre-stroke status. This change was identified using a –2 to +2 Lickert scale ranging from “much worse” to “much better” than before the stroke. We considered a domain not responsive for an ES <0.2, mildly responsive for an ES between 0.2 and 0.5, moderately responsive for an ES between 0.51 and 0.7 and highly responsive to change for an ES > 0.7 (Cohen, 1988).

Feasibility and acceptability were evaluated with the percentage of missing data and the distribution of responses by studying domain scores. We investigated the presence of floor or ceiling effects, defined by the proportion of patients who had a minimum or maximum score for a given domain. A score above 20% was considered high (Cruz-Cruz, Martinez-Nuñez, Perez, Kravzov-Jinich, Ríos-Castañeda, & Altagracia-Martinez, 2013).

2.4.1 Ethics

This study was approved by the ethics committee (Comité de Protection des Personnes Est 1, number N°: 2014-A01654-43). Patients were informed about the study and their consent was obtained at enrolment.

3 Results

3.1 Characteristics of patient

A total of 77 patients were included in this study. The baseline characteristics are shown in Table 1.

Table 1
Baseline characteristics of patients

n 95% CI

Age (years) (mean±SD) 64.3±15.4

Male gender 45 58.4% 46.9–69.1

Initial NIHSS score

Mean±SD 5.2±0.6

Median (IQR) 3 (2–7)

NIHSS score at inclusion

Mean±SD 1.6±2

Median (IQR) 1 (0–3)

Type of stoke

Cerebral infarction 74 96.1% 88,3–98,8

Intracranial hemorrhage 3 3.9% 1,2–11,7

Acute care

No acute treatment 51 68.2% 57,2–78,6

IV thrombolysis 20 27% 18–38,5

Associated IV thrombolysis and endovascular treatment 2 2.7% 0,7–10,5

Endovascular treatment only 1 1.4% 0,2–9,3

Barthel Index

Mean±SD 95.3±11.7

40 1 1.3% 0.2–9

50 1 1.3% 0.2–9

65 3 3.9% 1.2–11.7

80 3 3.9% 1.2–11.7

85 4 5.2% 1.9–13.3

90 1 1.3% 0.2–9

95 4 5.2% 1.9–13.3

100 60 77.9% 67–86

	n	95% CI
Age (years) (mean±SD)		64.3±15.4
Male gender	45	58.4%	46.9–69.1
Initial NIHSS score
Mean±SD		5.2±0.6
Median (IQR)		3 (2–7)
NIHSS score at inclusion
Mean±SD		1.6±2
Median (IQR)		1 (0–3)
Type of stoke
Cerebral infarction	74	96.1%	88,3–98,8
Intracranial hemorrhage	3	3.9%	1,2–11,7
Acute care
No acute treatment	51	68.2%	57,2–78,6
IV thrombolysis	20	27%	18–38,5
Associated IV thrombolysis and endovascular treatment	2	2.7%	0,7–10,5
Endovascular treatment only	1	1.4%	0,2–9,3
Barthel Index
Mean±SD		95.3±11.7
40	1	1.3%	0.2–9
50	1	1.3%	0.2–9
65	3	3.9%	1.2–11.7
80	3	3.9%	1.2–11.7
85	4	5.2%	1.9–13.3
90	1	1.3%	0.2–9
95	4	5.2%	1.9–13.3
100	60	77.9%	67–86

The mean age of patients was 64.3 years and 58.4% of them were men. The mean initial NIHSS score was 5.2±0.6. At the enrolment visit, the mean NIHSS score was 1.6±2, 67.6% of patients had a modified Rankin score of 0 to 1 (Fig. 1), and 83.1% had a Barthel index > 90 (Table 1).

Fig.1

Modified Rankin score in stroke patients three months after the inclusion visit.

3.2 Translation and cross-cultural adaptation of the SS-QoL

Small differences in semantics and sentence structure were observed between the original and the back-translated versions and between the French and the English versions. To obtain a consensus, every word and expression was discussed at the synthesis meeting to choose the best equivalence. During the pre-test, the sample of 10 patients had no missing responses. The items were clearly understood and none of the patients had problems with the wording of questions. The group of experts considered the questionnaire relevant and understandable. No changes were made after the pre-test.

3.3 Psychometric validity of the French version of the SS-QoL

Mean scores of the domains of the French version of the SS-QoL are presented in Table 2. The lowest scores were found for the Fatigue (3.2), Social Roles (3.4) and Personality (3.5) domains. Conversely, the Self-care (4.7), Upper extremity function (4.7) and Vision (4.5) domains had the highest mean scores. The distribution of the domain scores ranged from 1 to 5 except for the Language, Mobility, Mood, Energy, Thinking and Vision domains, whose lowest scores were 2.2 - 1.2 - 2 - 1.6 and 1.7, respectively.

Table 2
Internal consistency – floor and ceiling effects

SS-QOL Domain No. of items Mean (SD) Median Minimum Maximum score % Floor effect % Ceiling effect Cronbach’s α

Energy 3 3.2 (1.1) 3.3 1.0–5.0 2.6–7.8 0.69

Family roles 3 3.8 (1.2) 4.0 1.0–5.0 1.3–28.6 0.67

Language 5 4.4 (0.6) 4.6 2.2–5.0 0–24.7 0.82

Mobility 6 4.2 (0.9) 4.7 1.2–5.0 0–31.2 0.88

Mood 5 4.1 (0.9) 4.2 2.0–5.0 0–23.4 0.72

Personality 3 3.5 (1.1) 3.7 1.0–5.0 2.6–10.4 0.58

Self-care 5 4.7 (0.7) 5.0 1.6–5.0 0–77.9 0.91

Social roles 5 3.4 (1.1) 3.4 1.0–5.0 2.6–10.4 0.74

Thinking 3 3.7 (0.9) 3.7 1.7–5.0 0–13 0.65

Upper ext. function 5 4.5 (0.9) 5.0 1.0–5.0 1.3–50.7 0.89

Vision 3 4.7 (0.7) 5.0 2.0–5.0 0–66.2 0.65

Work/Productivity 3 3.8 (1.2) 4.0 1.0–5.0 3.9–26 0.80

SS-QOL Domain	No. of items	Mean (SD)	Median	Minimum Maximum score	% Floor effect % Ceiling effect	Cronbach’s α
Energy	3	3.2 (1.1)	3.3	1.0–5.0	2.6–7.8	0.69
Family roles	3	3.8 (1.2)	4.0	1.0–5.0	1.3–28.6	0.67
Language	5	4.4 (0.6)	4.6	2.2–5.0	0–24.7	0.82
Mobility	6	4.2 (0.9)	4.7	1.2–5.0	0–31.2	0.88
Mood	5	4.1 (0.9)	4.2	2.0–5.0	0–23.4	0.72
Personality	3	3.5 (1.1)	3.7	1.0–5.0	2.6–10.4	0.58
Self-care	5	4.7 (0.7)	5.0	1.6–5.0	0–77.9	0.91
Social roles	5	3.4 (1.1)	3.4	1.0–5.0	2.6–10.4	0.74
Thinking	3	3.7 (0.9)	3.7	1.7–5.0	0–13	0.65
Upper ext. function	5	4.5 (0.9)	5.0	1.0–5.0	1.3–50.7	0.89
Vision	3	4.7 (0.7)	5.0	2.0–5.0	0–66.2	0.65
Work/Productivity	3	3.8 (1.2)	4.0	1.0–5.0	3.9–26	0.80

3.3.1 Internal consistency

Cronbach’s alpha coefficients were satisfactory for seven domains (α> 0.7) and acceptable for four (α> 0.65), thus showing satisfactory homogeneity (Table 2). The internal consistency of the Personality domain was moderate (α= 0.58).

3.3.2 Factorial validity

The first 12 components of the factorial analysis explained 75% of the total variance of the items. This analysis did not exactly reconstitute the original structure of the scale (Table 3). Twelve groups of items organized into only eight factorial axes. Several theoretical domains overlapped on axis 1: Family role, Mobility, Self-care, Upper extremity function and Work/productivity domains gathered into one large domain. The Work/Productivity domain also overlapped on axis 4 with the Energy domain. Standardized loads were mostly higher than 0.5 (except for nine items), which indicated a good correlation between items and the domains to which they belonged.

Table 3
Exploratory factorial analysis with varimax rotation

F1 F2 F3 F4 F5 F6 F7 F8 F9 F10 F11 F12

Item 1 - en1 0.83

Item 2 - en2 0.49 0.42

Item 3 - en3 0.33 0.62 0.42

Item 4 - fr1 0.41 0.52 0.34

Item 5 - fr2 0.61

Item 6 - fr3 0.32 0.57

Item 7 - la1 0.85 0.31

Item 8 - la2 0.39 0.71

Item 9 - la3 0.84

Item 10 - la4 0.8

Item 11 - la5 0.84

Item 12 - mob1 0.74 0.53

Item 13 - mob2 0.59 0.56

Item 14 - mob3 0.64 0.49

Item 15 - mob4 0.68

Item 16 - mob5 0.76

Item 17 - mob6 0.58

Item 18 - mo1 0.70

Item 19 - mo2 0.33 0.69

Item 20 - mo3 0.30 0.62

Item 21 - mo4 0.67

Item 22 - mo5 0.68

Item 23 - pers1 0.33 0.60

Item 24 - pers2 0.65

Item 25 - pers3 0.67

Item 26 - sc1 0.77

Item 27 - sc2 0.80

Item 28 - sc3 0.90

Item 29 - sc4 0.94

Item 30 - sc5 0.67

Item 31 - sr1 0.82 0.39

Item 32 - sr2 0.86

Item 33 - sr3 0.72

Item 34 - sr4 0.35

Item 35 - sr5 0.35 0.32 0.49 0.67

Item 36 - th1 0.69

Item 37 - th2 0.50 0.65

Item 38 - th3 0.77

Item 39 - uf1 0.40 0.47

Item 40 - uf2 0.80

Item 41 - uf3 0.86

Item 42 - uf4 0.93

Item 43 - uf5 0.79 0.34

Item 44 - vi1 0.80

Item 45 - vi2 0.77

Item 46 - vi3 0.55 0.30

Item 47 - wp1 0.61 0.39 0.39

Item 48 - wp2 0.40 0.51

Item 49 - wp3 0.45 0.52 0.34

3.3.3 Construct validity - convergent validity

Table 4 shows the correlations between domains of the SS-QoL and the corresponding scales listed. The Energy, Family role, Work/Productivity, Mobility, Mood, Self-care and Upper extremity function domains correlated strongly with the FSS scale, the SF-36 (limitations due to physical condition - physical activity), the HAD scale, the Barthel index and the NIHSS score (motor arm), respectively (ρ> 0.5). The Vision, Language, Social Role domains correlated moderately with the Visual fields and Language domains of the NIHSS score and the SF-36 scale (ρ> 0.35). The Personality domain did not correlate with the HAD scale, nor did the Thinking dimension with the MMSE.

Table 4
External construct validity - convergent validity

SS-QOL Domain Established measure Spearman coefficient ρ p

Energy FSS –0.74 <0.0001

Family roles SF-36: emotional and physical role limitations 0.61 <0.0001

Language NIHSS: aphasia and dysarthria scores –0.38 0.001

Mobility SF-36: physical function subscale 0.81 <0.0001

Mood HAD: anxiety and depression subscales 0.56; 0.52 <0.0001

Personality HAD: anxiety and depression subscales –0.29; –0.26 0.01

Self-care Barthel index 0.89 <0.0001

Social roles SF-36: social functioning subscale 0.45 <0.0001

Thinking MMSE 0.29 0.01

Upper ext. function Barthel index: NIHSS: upper extremity score 0.74; –0.75 <0.0001

Vision NIHSS: visual field and ocular movements –0.48 <0.0001

Work/Productivity SF-36: physical role limitations subscale 0.57 <0.0001

SS-QOL Domain	Established measure	Spearman coefficient ρ	p
Energy	FSS	–0.74	<0.0001
Family roles	SF-36: emotional and physical role limitations	0.61	<0.0001
Language	NIHSS: aphasia and dysarthria scores	–0.38	0.001
Mobility	SF-36: physical function subscale	0.81	<0.0001
Mood	HAD: anxiety and depression subscales	0.56; 0.52	<0.0001
Personality	HAD: anxiety and depression subscales	–0.29; –0.26	0.01
Self-care	Barthel index	0.89	<0.0001
Social roles	SF-36: social functioning subscale	0.45	<0.0001
Thinking	MMSE	0.29	0.01
Upper ext. function	Barthel index: NIHSS: upper extremity score	0.74; –0.75	<0.0001
Vision	NIHSS: visual field and ocular movements	–0.48	<0.0001
Work/Productivity	SF-36: physical role limitations subscale	0.57	<0.0001

3.3.4 Construct validity - divergent validity

The results of the analysis of known-group comparisons are presented in Tables 5 and 6. Nine domain scores of the 12 were significantly different between patients reporting a worse or better quality of life compared with their pre-stroke status (p < 0.05). For three domains (Thinking, Personality, Language) the difference was not significant. Furthermore, the total score for the French version of the SS-QoL scale was significantly different between these two groups (4.5 versus 3.8; p < 0.001) (Table 5). Similarly, the total score for the French version of the SS-QoL scale was significantly different between patients with an initial Rankin score less than or equal to 1 and patients whose Rankin score was higher than 1 (4.3 vs. 3.6, p < 0.001) (Table 6).

Table 5
External construct validity - divergent validity

SS-QOL domain Group 1 n = 25 Mean (CI) Group 2 n = 52 Mean (CI) p

Energy 4 (3.3–4.7) 3 (2–3.7) <0.001

Family Roles 5 (4.3–5.0) 3.3 (2.3–4.2) <0.001

Language 4.6 (4.2–4.8) 4.6 (4.2–5) 0.71

Mobility 5 (4.7–5) 4.2 (3.1–4.8) <0.001

Mood 4.6 (4–5) 4 (3.5–4.7) 0.028

Personality 4 (3–4.7) 3.3 (2.7–4) 0.07

Self-care 5 (5–5) 5 (4.5–5) 0.03

Social roles 4 (3–4.8) 3.1 (2.2–4.2) 0.01

Thinking 4 (3–4.7) 3.5 (2.8–4.7) 0.18

Upper ext function 5 (4.8–5) 4.6 (3.9–5) 0.003

Vision 5 (5–5) 5 (4.5–5) 0.032

Work/Productivity 4.7 (4.3–5) 3.7 (2.7–4.3) <0.001

Total 4.5 (4.2–4.6) 3.8 (3.5–4.2) <0.001

SS-QOL domain	Group 1 n = 25 Mean (CI)	Group 2 n = 52 Mean (CI)	p
Energy	4 (3.3–4.7)	3 (2–3.7)	<0.001
Family Roles	5 (4.3–5.0)	3.3 (2.3–4.2)	<0.001
Language	4.6 (4.2–4.8)	4.6 (4.2–5)	0.71
Mobility	5 (4.7–5)	4.2 (3.1–4.8)	<0.001
Mood	4.6 (4–5)	4 (3.5–4.7)	0.028
Personality	4 (3–4.7)	3.3 (2.7–4)	0.07
Self-care	5 (5–5)	5 (4.5–5)	0.03
Social roles	4 (3–4.8)	3.1 (2.2–4.2)	0.01
Thinking	4 (3–4.7)	3.5 (2.8–4.7)	0.18
Upper ext function	5 (4.8–5)	4.6 (3.9–5)	0.003
Vision	5 (5–5)	5 (4.5–5)	0.032
Work/Productivity	4.7 (4.3–5)	3.7 (2.7–4.3)	<0.001
Total	4.5 (4.2–4.6)	3.8 (3.5–4.2)	<0.001

Quality of life at baseline compared with pre-stroke status. Group 1 = not worse QOL; Group 2 = worse QOL.

Table 6

External construct validity: divergent validity

SS-QOL domain	Group 1 n = 25 Mean (CI)	Group 2 n = 52 Mean (CI)	p
Energy	2.7 (2–3)	3.7 (2.7–4.5)	0.002
Family role	3 (2.3–3.7)	4.3 (4–5)	<0.001
Language	4.6 (4–4.8)	4.6 (4.2–5)	0.4
Mobility	3.2 (2.7–4.2)	4.8 (4.3–5)	<0.001
Mood	4 (3.6–4.6)	4.4 (3.8–4.9)	0.15
Personality	4 (2.7–4.3)	3.7 (2.7–4.2)	0.68
Self-care	4.2 (3.4–5)	5 (5–5)	<0.01
Social roles	2.8 (1.8–3.4)	4 (3–4.5)	0.003
Thinking	4 (2.7–4.7)	3.7 (3–4.7)	0.7
Upper ext function	4 (3.2–4.4)	5 (4.8–5)	<0.001
Vision	5 (4.3–5)	5 (4.7–5)	0.91
Work/Productivity	2.7 (2.3–3.3)	4.7 (4–5)	<0.001
Total	3.6 (3.2–3.8)	4.3 (4.1–4.5)	<0.001

Quality of life according to the modified Rankin score. Group 1: mRS > 1; Group 2: mRS = (0-1).

3.3.5 Reliability

All domains of the SS-QoL had excellent test-retest and inter-observer reliability. Intraclass correlation coefficients were all above 0.88 (Table 7).

Table 7
Fidelity

Test-retest n = 31 Inter observer n = 30

SS-QOL Domain ICC p ICC p

Energy 0.97 <0.001 0.95 <0.001

Family roles 0.96 <0.001 0.96 <0.001

Language 0.91 <0.001 0.88 <0.001

Mobility 0.97 <0.001 0.96 <0.001

Mood 0.97 <0.001 0.97 <0.001

Personality 0.98 <0.001 0.92 <0.001

Self-care 0.99 <0.001 0.99 <0.001

Social roles 0.97 <0.001 0.97 <0.001

Thinking 0.95 <0.001 0.91 <0.001

Upper ext. function 0.98 <0.001 0.97 <0.001

Vision 0.98 <0.001 0.97 <0.001

Work/Productivity 0.95 <0.001 0.96 <0.001

	Test-retest n = 31	Inter observer n = 30
Energy	0.97	<0.001	0.95	<0.001
Family roles	0.96	<0.001	0.96	<0.001
Language	0.91	<0.001	0.88	<0.001
Mobility	0.97	<0.001	0.96	<0.001
Mood	0.97	<0.001	0.97	<0.001
Personality	0.98	<0.001	0.92	<0.001
Self-care	0.99	<0.001	0.99	<0.001
Social roles	0.97	<0.001	0.97	<0.001
Thinking	0.95	<0.001	0.91	<0.001
Upper ext. function	0.98	<0.001	0.97	<0.001
Vision	0.98	<0.001	0.97	<0.001
Work/Productivity	0.95	<0.001	0.96	<0.001

3.3.6 Responsiveness

Responsiveness was assessed in each domain for patients who had reported a change in their quality of life between inclusion and their pre-stroke status, using a –2 to +2 Lickert scale ranging from “much worse” to “much better”. Three patients were lost because of one death and two refusals to participate. Nine of the twelve domains of the French version of the SSQOL were mildly to moderately responsive (ES> 0.3) (Table 8). The Energy domain was the most responsive (ES> 0.7). In contrast, Social Role and Work/Productivity domains were the least responsive (ES< 0.2).

Table 8
Responsiveness

SS-QOL Domain Mean change (SD) SRM Effect size n affected

Energy –0.01 (0.76) –0.02 1.04 57

Family roles 0.28 (0.97) 0.29 0.33 19

Language 0.44 (0.59) 0.75 0.65 29

Mobility 0.25 (0.74) 0.34 0.26 34

Mood 0.25 (0.80) 0.33 0.30 32

Personality 0.38 (1.14) 0.33 0.46 30

Self-care 0.52 (0.6) 0.86 0.49 16

Social roles 0.04 (1.04) 0.04 0.04 15

Thinking 0.23 (0.74) 0.31 0.29 40

Upper ext. function 0.35 (0.53) 0.66 0.32 29

Vision 0.47 (0.55) 0.86 0.52 19

Work/Productivity 0.15 (0.72) 0.2 0.14 42

SS-QOL Domain	Mean change (SD)	SRM	Effect size	n affected
Energy	–0.01 (0.76)	–0.02	1.04	57
Family roles	0.28 (0.97)	0.29	0.33	19
Language	0.44 (0.59)	0.75	0.65	29
Mobility	0.25 (0.74)	0.34	0.26	34
Mood	0.25 (0.80)	0.33	0.30	32
Personality	0.38 (1.14)	0.33	0.46	30
Self-care	0.52 (0.6)	0.86	0.49	16
Social roles	0.04 (1.04)	0.04	0.04	15
Thinking	0.23 (0.74)	0.31	0.29	40
Upper ext. function	0.35 (0.53)	0.66	0.32	29
Vision	0.47 (0.55)	0.86	0.52	19
Work/Productivity	0.15 (0.72)	0.2	0.14	42

3.3.7 Feasibility and acceptability

The acceptability of our questionnaire was good and there were no missing responses. The distribution of responses reflected no floor effect (Table 2). We found a ceiling effect ranging from 23.4% to 66.2% for seven domains (Family role, Language, Mobility, Mood, Self-care, Upper extremity function, Vision and Work/Productivity). It took a mean of 10.3 minutes to complete the SS-QoL.

4 Discussion

This study demonstrated that the French version of the SS-QoL scale appears to be reliable, valid, and moderately responsive to change to assess quality of life in stroke patients.

The feasibility and acceptability of the SS-QoL scale were good. There were neither missing data nor floor effects. Although there were moderate ceiling effects in some areas, only the Vision domain had a significant ceiling effect (66.2%), which is consistent with findings from the validation of both the original scale (Williams, et al., 1999), and its Danish version (Muus, Williams, & Ringsberg, 2007).

Seven of the twelve domains showed good internal consistency with α> 0.7, and four areas showed acceptable internal consistency with α> 0.65 (Energy, Family role, Thinking and Vision). In contrast, the Personality domain showed the lowest internal consistency (α= 0.58). Validation studies of other SS-QoL versions did not report such a low internal consistency for the Personality domain. Indeed, the lowest reported α was for the Thinking domain in the Turkish version (Hakverdioğlu Yönt, & Khorshid, 2012), and for the Family role domain in the Mexican version (Cruz-Cruz, et al., 2013). Our result may indicate poor homogeneity of the Personality domain. One hypothesis could be that the different items contained in this domain measure several concepts (evolution of the personality since the stroke or fixed aspects of personality). Another explanation could be an issue of translation and cross-cultural adaptation into French. However, after reviewing the different stages of the translation, the wording of questions did not appear to be problematic. Finally, it cannot be excluded that the low internal consistency for the Personality domain could be attributable to a sampling bias. Indeed, Cronbach’s alpha coefficient depends on the total number of items and on the size and the heterogeneity of the sample. As a consequence, a small and homogeneous group of patients decreases the α (Fayers, & Machin, 2000). Because our study was population-based, initial stroke severity was low (mean NIHSS of 5) as compared with hospital-based recruitment, and the functional outcomes of patients at 3 months were therefore good. As a result, this small sample appeared to be quite homogeneous, which may explain lower overall α values than those reported for the original (α= 0.73 to 0.89) and Danish (α= 0.81 to 0.94) versions of the SS-QoL.

Exploratory factorial analysis showed that the French version of the SS-QoL was organized into eight factorial axes. Items from the Family role, Mobility, Self-care, Upper-extremity function and Work/productivity domains seemed to gather into a new large domain. The Turkish (Hakverdioğlu Yönt, & Khorshid, 2012) and German (Ewert, & Stucki 2007) versions also highlighted a reduced number of domains, with a large domain gathering Self-care, Work/productivity, Mobility and Upper-extremity function dimensions. In the original development of the SS-QoL, the authors suggested that a smaller number of domains could be found in future factorial analyses (Williams, et al., 1999). They especially pointed out that the exact number of domains was less important than the content of items, as the final goal was to accurately assess all of the potential effects of stroke, in a patient-centered way. Although our factorial structure did not exactly correspond to the original version, the overall consistency of the scale was maintained.

The French version of the SS-QoL scale showed good convergent validity. Seven domains (Fatigue, Family role, Work/Productivity, Mobility, Mood, Autonomy, Functionality of the upper limbs) correlated strongly with validated external associated scales (ρ> 0.5). The Thinking domain did not correlate with the MMSE. This expected result could be explained by the fact that the MMSE is not sensitive enough to screen for cognitive impairment after stroke (Sivakumar, Kate, Jeerakathil, Camicioli, Buck, & Butcher, 2014; Pendlebury, Mariz, Bull, Mehta, & Rothwell, 2012), especially to explore dysexecutive syndromes. In addition, patients with cognitive impairment were excluded from the study because of their inability to complete the questionnaire. In contrast, items from the Thinking domain of the SS-QoL were very sensitive, and were better able to detect subtle functional signs (for example: “do you have trouble remembering things?”). The Personality domain did not correlate with the HAD. This result was not found in either the original or in the Danish version, in which the Beck Depression Inventory was used as the comparative tool (Beck, Ward, Mendelson, Mock, & Erbaugh, 1960). This finding suggests that items contained in the Personality domain explored different aspects from those studied in the HAD scale, i.e. more emotional elements related to the patient’s personality rather than anxiety and depression aspects. As observed in the original, Danish and Mexican versions, the Vision and Language domains correlated moderately with corresponding items of the NIHSS (ρ> 0.35). The NIHSS score is not a powerful tool to assess these clinical features exhaustively. Actually, it measures neurological impairment rather than disability; it thus underestimates the impact of a neurological symptom on QOL. The use of a specific language test and a systematic visual field examination would have overcome this limitation but was not applicable in this study.

In support of its divergent validity, the French version of the SS-QoL scale was able to differentiate groups with known differences. Scores from patients who reported a worse quality of life compared with their pre-stroke status significantly differed from those who did not, which is consistent with a previous report (Williams, et al., 1999; Muus, Williams, & Ringsberg, 2007). Similarly, using the mRS as an objective endpoint, the overall score was significantly different between patients with an mRS > 1 at inclusion and those with an mRS equal to 0 or 1.

Other positive psychometric properties were the excellent test-retest and inter-observer reliabilities for a sample of 60 patients (ICC> 0.88). Concerning responsiveness, most domains of the French version of the SS-QoL were mildly to moderately responsive, but two domains were not responsive to change: Work/Productivity and Social role. Two studies previously investigated the responsiveness of the SS-QoL. In the Danish version, based on analyses of effect size, the authors concluded that eight domains were mildly to moderately responsive (SRM = 0.33–0.56) and four were non-responsive (Muus, Christensen, Petzold, Harder, Johnsen, Kirkevold, & Ringsberg, 2011). In contrast, better results were demonstrated in the original version (Williams, et al., 1999). The use of a homogeneous sample of patients with low initial severity and good functional outcomes in our study may explain our findings, because of ceiling effects, thus limiting any further evaluation of responsiveness. Another hypothesis could be that the time between the two measurements was too short.

Psychometric properties of different scales are difficult to compare because of their heterogeneity and the extended range of statistical analysis. The SS-QoL can be compared with the SIS, that is the only other specific scale validated in French. For content validity, we noted that in the validation study for the SIS, about 25% of patients responded that they would have liked to be asked about sexuality and fatigue (Caël, et al., 2015). Despite a smaller number of items, the SS-QoL explored a wider range of domains by including areas such as Fatigue, Personality and Vision. Moreover, one SS-QoL item does assess sexuality issues. The SS-QoL test-retest and inter-observer reliabilities were better with ICC all above 0.88 versus 0.65 in the SIS test-retest. Both scales suffered from ceiling effects. As expected and previously discussed, except for the Personality domain, α coefficients ranged from 0.65 to 0.91 in the SS-QoL versus 0.89 to 0.97 in the SIS. Both scales showed good divergent validity according to different endpoints: professional activity, Boston Diagnostic Aphasia Examination, NIHSS and BI scores in the SIS, mRS and self-declared overall quality of life in the SS-QoL. Convergent validity was more difficult to analyse because SIS scores were compared with those from the BI, the HAD and the Duke profile. Conversely, SS-QoL domain scores correlated with SF 36, HAD, BI, NIHSS, MMSE and FSS scores. In both scales, Self-care and Mood domains correlated well with the BI and the HAD, respectively (ρ= –0.56 in SIS versus 0.56 in SS-QoL and ρ= 0.89 in SS-QoL versus 0.71 the SIS, respectively). Concerning responsiveness, the SRM range was quite similar in both scales: 0.2–0.86 in the SS-QoL (except for the Energy domain which was not explored in the SIS) versus 0.16 to 0.9 in the SIS. The SRM of some domains, such as the Self-care domain (SRM = 0.8 in SIS versus 0.86 in SS-QoL) and the Hand function domain (SRM = 0.57 in SIS versus 0.66 in SS-QoL), were very similar in both scales. The SRM of the Language domain was better in the SS-QoL (0.75 versus 0.26) whereas the SRM of the Mobility domain was higher in the SIS (0.34 versus 0.71).

Our study has several limitations. As discussed above, the relatively small study sample may have influenced the internal consistency of the scale. Similarly, the homogeneity of our patients may have limited its psychometric properties. For example, two thirds of our patients had a low residual handicap (mRS < 2), which is not representative of the general stroke population. This situation can be explained by the fact that we excluded patients with a history of previous symptomatic stroke, and patients who could not be tested because they had severe aphasia or dementia, or were bedridden. However, because of the population-based setting of our study, patients were representative of those seen during follow-up in clinical practice, thus reinforcing the usefulness of this tool by clinicians.

5 Conclusion

The French version of the SS-QoL scale has satisfactory psychometric properties and thus appears to be a useful tool for measuring QOL in stroke patients. QOL is an emergent outcome that will be addressed in future epidemiological studies and clinical trials.

Conflict of interest

Yannick Béjot received honoraria or consulting fees from AstraZeneca France, Daiichi-Sankyo, Pfizer, BMS, Covidien, Bayer and MSD France. Other authors: none.

References

Ali

, Fulton

, Quinn

, & Brady

(2013). How well do standard stroke outcome measures reflect quality of life? A retrospective analysis of clinical trial data. Stroke, 44, 3161–3165.

Beck

A. T.

, Ward

C. H.

, Mendelson

, Mock

, & Erbaugh

(1960). An inventory for measuring depression. Arch Gen Psychiatry, 4, 53–63.

Béjot

, Rouaud

, Durier

, Caillier

, Marie

, Freysz

, Yeguiayan

J. M.

, Chantegret

, Osseby

, Moreau

, & Giroud

(2007). Decrease in the stroke case fatality rates in a French population-based twenty-year study. A comparison between men and women. Cerebrovasc Dis, 24, 439–444.

Caël

, Decavel

, Binquet

, Benaim

, Puyraveau

, Chotard

, Moulin

, Parratte

, Béjot

, & Mercier

(2015). Stroke impact scale version 2: Validation of the French version. Phys Ther, 95, 778–790.

Cohen

(1988). Statistical power analysis for the behavioral sciences. Second edition. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Cruz-Cruz

, Martinez-Nuñez

J. M.

, Perez

M. E.

, Kravzov-Jinich

, Ríos-Castañeda

, & Altagracia-Martinez

(2013). Evaluation of the Stroke-Specific Quality-of-Life (SSQOL) Scale in Mexico: A preliminary approach. Value in Health Regional Issues, 2, 392–397.

de Perreti

, Grimaud

, Tuppin

, Chin

, & Woimant

(2012). Prevalence of stroke, sequelae and difficulties in activities of daily living: French Disabilities and Health studies 2008-2009. BEH, 1, 1–6.

Devilliers

, Amoura

, Besancenot

J. F.

, Bonnotte

, Pasquali

J. L.

, Wahl

, Maurier

, Kaminsky

, Pennaforte

J. L.

, Magy-Bertrand

, Arnaud

, Binquet

, & Guillemin

(2012). LupusQoL-FR is valid to assess quality of life in patients with systemic lupus erythematosus. Rheumatology (Oxford), 51, 1906–1915.

Epstein

, Santo

R. M.

, & Guillemin

(2015). A review of guidelines for cross-cultural adaptation of questionnaires could not bring out a consensus. J Clin Epidemiol, 68, 435–441.

10.

Ewert

, & Stucki

(2007). Validity of the SS-QOL in Germany and in survivors of hemorrhagic or ischemic stroke. Neurorehabil Neural Repair, 21, 161–168.

11.

Fayers

, & Machin

(2000). Quality of Life: Assessment, Analysis and Interpretation. First edition. John Wiley & Sons.

12.

Hakverdioğlu Yönt

, & Khorshid

(2012). Turkish version of the Stroke-Specific Quality of Life Scale. Int Nurs Rev, 59, 274–280.

13.

Koton

, Schneider

A. L.

, Rosamond

W. D.

, Shahar

, Sang

, Gottesman

R. F.

, & Coresh

(2014). Stroke incidence and mortality trends in US communities, 1987 to 2011. JAMA, 312, 259–268.

14.

Kunst

A. E.

, Amiri

, & Janssen

(2011). The decline in stroke mortality: Exploration of future trends in 7 Western European countries. Stroke, 42, 2126–2130.

15.

Lance

C. E.

, Butts

M. M.

, & Michels

L. C.

(2006). The Sources of Four Commonly Reported Cutoff Criteria: What Did They Really Say? Organizational Research Methods, 9, 202–220.

16.

Miller

E. L.

, Murray

, Richards

, Zorowitz

R. D.

, Bakas

, Clark

, & Billinger

S. A.

; American Heart Association Council on Cardiovascular Nursing and the Stroke Council (2010). Comprehensive overview of nursing and interdisciplinary rehabilitation care of the stroke patient: A scientific statement from the American Heart Association. Stroke, 41, 2402–2448.

17.

Mokkink

L. B.

, Terwee

C. B.

, Patrick

D. L.

, Alonso

, Stratford

P. W.

, Knol

D. L.

, Bouter

L. M.

, & de Vet

H. C.

(2010). The COSMIN study reached international consensus on taxonomy, terminology, and definitions of measurement properties for health- related patient-reported outcomes. J Clin Epidemiol, 63, 737–745.

18.

Muus

, Williams

L. S.

, & Ringsberg

K. C.

(2007). Validation of the Stroke Specific Quality of Life Scale (SS-QOL): Test of reliability and validity of the Danish version (SS-QOL-DK). Clinical Rehabilitation, 21, 620–627.

19.

Muus

, Christensen

, Petzold

, Harder

, Johnsen

S. P.

, Kirkevold

, & Ringsberg

K. C.

(2011). Responsiveness and sensitivity of the Stroke Specific Quality of Life Scale Danish version. Disabil Rehabil, 33, 2425–2433.

20.

Nunally

J. C.

(1978). Psychometric Theory. Second edition. New York: McGraw-Hill.

21.

Ojo Owolabi

(2010). Psychometric properties of the HRQOLISP-40: A novel, shortened, multiculturally, valid holistic stroke measure. Neurorehabil Neural Repair, 24, 814–825.

22.

Pendlebury

S. T.

, Mariz

, Bull

, Mehta

, & Rothwell

P. M.

(2012). MoCA, ACE-R, and MMSE Versus the National Institute of Neurological Disorders and Stroke – Canadian Stroke Network Vascular Cognitive Impairment Harmonization Standards Neuropsychological Battery After TIA and Stroke. Stroke, 43, 464–469.

23.

Polit

D. F.

(1996). Data analysis and statistics for nursing research. First edition. Appleton & Lange.

24.

Sivakumar

, Kate

, Jeerakathil

, Camicioli

, Buck

, & Butcher

(2014). Serial Montreal Cognitive Assessments Demonstrate Reversible Cognitive Impairment in Patients With Acute Transient Ischemic Attack and Minor Stroke. Stroke, 45, 1709–1715.

25.

Streiner

D. L.

, & Norman

G. R.

(1995). Health Measurement Scales: A Practical Guide to Their Development and Use. Second edition. Oxford University Press.

26.

Van Straten

, De Haan

R. J.

, Limburg

, Schuling

, Bossuyt

P. M.

, & van den Bos

G. A.

(1997). A stroke-adapted 30-item version of the Sickness Impact Profile to assess quality of life (SA-SIP30). Stroke, 28, 2155–2161.

27.

World Health Organization. (1993). Study protocol for the World Health Organization project to develop a quality of life assessment instrument (WHOQOL). Qual Life Res, 2, 153–159.

28.

Williams

L. S.

, Weinberger

, Harris

L. E.

, Clark

D. O.

, & Biller

(1999). Development of a stroke-specific quality of life scale. Stroke, 30, 1362–1369.

	F1	F2	F3	F4	F5	F6	F7	F8	F9	F10	F11	F12
Item 1 - en1				0.83
Item 2 - en2				0.49	0.42
Item 3 - en3	0.33			0.62		0.42
Item 4 - fr1	0.41	0.52		0.34
Item 5 - fr2	0.61
Item 6 - fr3	0.32								0.57
Item 7 - la1			0.85				0.31
Item 8 - la2	0.39		0.71
Item 9 - la3			0.84
Item 10 - la4							0.8
Item 11 - la5			0.84
Item 12 - mob1	0.74				0.53
Item 13 - mob2	0.59				0.56
Item 14 - mob3	0.64				0.49
Item 15 - mob4					0.68
Item 16 - mob5	0.76
Item 17 - mob6	0.58
Item 18 - mo1						0.70
Item 19 - mo2		0.33								0.69
Item 20 - mo3						0.30				0.62
Item 21 - mo4						0.67
Item 22 - mo5						0.68
Item 23 - pers1			0.33						0.60
Item 24 - pers2									0.65
Item 25 - pers3									0.67
Item 26 - sc1	0.77
Item 27 - sc2	0.80
Item 28 - sc3	0.90
Item 29 - sc4	0.94
Item 30 - sc5	0.67
Item 31 - sr1		0.82										0.39
Item 32 - sr2		0.86
Item 33 - sr3		0.72
Item 34 - sr4												0.35
Item 35 - sr5	0.35	0.32			0.49							0.67
Item 36 - th1											0.69
Item 37 - th2							0.50				0.65
Item 38 - th3							0.77
Item 39 - uf1	0.40			0.47
Item 40 - uf2	0.80
Item 41 - uf3	0.86
Item 42 - uf4	0.93
Item 43 - uf5	0.79					0.34
Item 44 - vi1								0.80
Item 45 - vi2								0.77
Item 46 - vi3								0.55			0.30
Item 47 - wp1	0.61			0.39	0.39
Item 48 - wp2	0.40			0.51
Item 49 - wp3	0.45			0.52	0.34