Introduction
Multiple sclerosis (MS) is a chronic autoimmune disease characterized by inflammatory demyelination and neurodegeneration within the central nervous system, primarily affecting young adults.1 The disease presents with a wide range of neurological signs and symptoms that frequently lead to progressive physical and cognitive disability over time.2 As a result, rehabilitation has become a fundamental component of comprehensive MS care, aiming to preserve functional capacity, promote independence, and improve quality of life.3 In Spain, both the prevalence and incidence of MS have shown a marked increase in recent decades, positioning the country within a moderate-to-high risk area for the disease.4 Regular physical activity (PA) and structured exercise are widely recognized as essential components of MS rehabilitation. Exercise has been consistently associated with improvements in physical fitness, mobility, fatigue, and overall health outcomes, while reducing the risk of secondary complications related to physical inactivity and sedentary behaviour.5,6 Despite this evidence, people with MS (PwMS) remain less physically active and spend more time engaged in sedentary behavior than the general population, regardless of disability level.7 This pattern has also been confirmed in Spanish PwMS, among whom low levels of PA and high rates of sedentary behaviour have been reported, a situation that was further exacerbated during the COVID-19 pandemic.8 Sedentary behaviour has been associated with greater disability, reduced aerobic capacity, impaired balance and muscle strength, and an increased risk of comorbidities, all of which may further compromise functional independence.9–12
Although the benefits of exercise are well established, promoting long-term adherence remains a major challenge in MS rehabilitation. Adherence to exercise is influenced not only by physical limitations but also by psychological, social, and environmental factors that affect an individual’s willingness and ability to engage in regular PA.13 Among these determinants, self-efficacy has consistently emerged as one of the strongest predictors of health-related behaviors, including exercise participation.14
Bandura defined self-efficacy as an individual’s belief in their capability to organize and execute the actions required to achieve specific goals or manage challenging situations.15 In chronic diseases such as MS, self-efficacy influences how individuals perceive their condition, cope with symptoms, overcome barriers, and maintain health-promoting behaviours.16–19 Because MS is characterized by an unpredictable disease course and fluctuating symptoms, many PwMS experience reduced confidence in their ability to exercise regularly. Consequently, lower exercise self-efficacy may contribute to reduced participation in PA and poorer adherence to rehabilitation programs, particularly when exercise is perceived as difficult or uncomfortable.20
Given its role in determining whether individuals initiate and maintain exercise despite symptoms and other barriers, exercise self-efficacy can be considered a potentially modifiable psychological self-efficacy determinant of exercise adherence. Individuals with higher exercise self-efficacy are more likely to initiate, maintain, and resume exercise after setbacks, ultimately supporting long-term improvements in physical fitness and health outcomes.21 Reliable assessment of exercise self-efficacy is therefore essential for identifying patients at risk of poor adherence, tailoring rehabilitation interventions, and evaluating treatment outcomes through patient-reported outcome measures (PROMs). Currently, two instruments have been used to assess exercise self-efficacy in the context of MS: the Exercise Self-Efficacy Scale (EXSE-McAuley)22 and the Exercise Self-Efficacy (ESE-Neupert).23
However, the use of a PROM developed in a language or cultural context different from that of the target population requires more than a simple linguistic translation. Cross-cultural adaptation is necessary to ensure that the translated instrument preserves the conceptual meaning and measurement equivalence of the original scale and that its items are understandable, relevant, and appropriate for the target population.
Despite their potential clinical usefulness, neither instrument has previously been translated and culturally adapted into Spanish, and evidence regarding their measurement properties in Spanish-speaking PwMS is lacking. Therefore, the aim of this study was to translate and culturally adapt the Exercise Self-Efficacy Scale (EXSE-McAuley) and the Exercise Self-Efficacy (ESE-Neupert) questionnaire into Spanish, and to evaluate their measurement properties in Spanish PwMS.
Materials and Methods
Study Design
A methodological study of cross-cultural adaptation and evaluation of measurement properties, including a longitudinal test–retest component, was conducted. The study design, conduct, and reporting followed the Consensus-based Standards for the Selection of Health Measurement Instruments (COSMIN) guidelines.24
Participants, Collected Data and Sample Size
Participants were recruited consecutively from among patients undergoing a four-month outpatient multidisciplinary rehabilitation program at the Multiple Sclerosis Center of Catalonia (Cemcat, Barcelona, Spain) between May 2023 and January 2024. This program consisted of individual and group physiotherapy sessions delivered 2–3 times per week, based on therapeutic goals agreed upon with each patient.
Eligible participants were adults aged 18–69 years with a confirmed diagnosis of MS according to the 2017 McDonald criteria,25 who had been relapse-free for at least 30 days and had an Expanded Disability Status Scale (EDSS) score below 8.5. All participants underwent a comprehensive neuropsychological assessment by a neuropsychologist, including the Brief International Cognitive Assessment for Multiple Sclerosis (BICAMS)26 to establish their cognitive profile. Individuals with neurological disorders other than MS or with cognitive impairment precluding adequate completion of the questionnaires, as determined by this assessment, were excluded.
Sociodemographic and clinical variables (including age, gender, MS phenotype, disease duration and disability level) were collected by the multidisciplinary rehabilitation team at the start of the rehabilitation program, as part of the standard clinical assessment, and recorded in the shared clinical chart. These variables were therefore retrospectively extracted from the medical record for the purposes of the present study.
Sample size was estimated according to Zou et al27 Assuming an expected intraclass correlation coefficient (ICC) of 0.90, a minimum acceptable ICC of 0.75, a significance level of 5%, statistical power of 80%, and an anticipated 10% attrition rate, a minimum sample of 87 participants was required. This estimate also fulfilled COSMIN recommendations for validation studies, which suggest including 5–10 participants per questionnaire item.28
Instruments
The Exercise Self-Efficacy Scale (EXSE), developed by McAuley,22 consists of eight items assessing an individual’s confidence in maintaining moderate-intensity exercise three times per week for 40 minutes per session, over eight consecutive weeks, with each item corresponding to a successive week. Each item is scored from 0 (“not at all confident”) to 100 (“highly confident”), and the overall score corresponds to the mean of all items, with higher scores indicating greater exercise self-efficacy. The original EXSE-McAuley demonstrated excellent internal consistency (Cronbach’s α = 0.96) and was shown to be a good predictor of exercise maintenance.29,30
The Exercise Self-Efficacy (ESE), developed by Neupert et al,23 consists of nine items evaluating confidence in performing exercise under challenging circumstances, such as fatigue or competing daily demands. Responses are rated on a four-point Likert scale ranging from 1 (“not at all sure”) to 4 (“very sure”), with higher mean scores reflecting greater exercise self-efficacy. This scale was found to have internal consistency reliability, using Cronbach’s alpha = 0.88 in the original validation study.23
Translation and Cross-Cultural Adaptation
Translation and cross-cultural adaptation of both questionnaires followed the guidelines proposed by Beaton et al.31
Two independent bilingual translators performed the forward translation from English into Spanish. After reconciliation into a single version, an independent translator blinded to the original questionnaires performed the back-translation. The original and back-translated versions were compared to identify conceptual discrepancies, which were resolved by consensus among the research team.
The pre-final Spanish versions (EXSE-Sp and ESE-Sp) were pilot-tested in two PwMS to assess clarity, comprehensibility, and cultural relevance. No difficulties in understanding or completing the questionnaires were identified, and both instruments were considered suitable for use in the Spanish MS population. The final Spanish versions are provided in Supplementary Material 1.
Data Analysis
Statistical analyses were performed using IBM SPSS Statistics version 28.0 (IBM Corp., Armonk, NY, USA) and Winsteps software. Statistical significance was established at p < 0.05. Continuous variables are presented as means and standard deviations or medians and interquartile ranges, as appropriate.
Categorical variables are presented as frequencies and percentages. Measurement properties were analyzed according to the procedures and criteria described in the subsections below. Missing data were minimal and were handled using pairwise deletion.
Measurement Properties
Measurement properties were evaluated according to COSMIN recommendations28 using a Classical Test Theory (CTT) framework. The CTT framework was considered the most appropriate primary analytical approach based on the study objectives and the conceptual structure of both scales. The Exercise Self-Efficacy Scale (ESE; Neupert et al, 2009) does not imply a clear hierarchical ordering of the items.
In the Exercise Self-Efficacy Scale (EXSE; McAuley et al, 1993), although the item structure could suggest an ordinal progression, responses do not necessarily follow a hierarchical order. Therefore, the conceptual model and structure of both scales were considered more consistent with a CTT framework for evaluating their measurement properties. As a complementary exploratory analysis, a polytomous Rasch analysis was conducted using Winsteps software32 to examine item-level functioning.
The analysis included evaluation of item fit using Infit and Outfit mean-square (MNSQ) with values >1.5 considered indicative of potential misfit (Supplementary Table 1 and Supplementary Table 2). Point-measure correlations were also examined to assess the contribution of individual items to the underlying trait. Detailed results are provided in Supplementary Material 2.
Descriptive analysis per item and scale:
Floor and Ceiling Effects
Floor and ceiling effects were evaluated according to Terwee et al,33 considering values <5% as absent, 5–15% as mild, >15% as present, and >30% as severe.
Reliability
Internal Consistency Reliability
Internal consistency was assessed using Cronbach’s alpha.34 Values between 0.70 and 0.90 were considered indicative of acceptable internal consistency according to COSMIN recommendations.35
Test–Retest Reliability
Test–retest reliability was assessed using a two-way mixed-effects model, absolute agreement, single measures intraclass correlation coefficient (ICC), with 95% confidence intervals. Reliability was interpreted according to Fleiss,36 with ICC values ≥0.80 considered excellent, 0.60–0.79 good, 0.40–0.59 moderate, and <0.40 poor.
Participants completed both questionnaires twice during the final two weeks of their four-month rehabilitation program, with an interval of 7–10 days between assessments. Participants were instructed to maintain their usual exercise routine during this period, following the procedure described in the original EXSE-McAuley study.22 The test–retest interval was scheduled during the final two weeks of the participants’ four-month rehabilitation program. At this advanced stage of the program, exercise self-efficacy was considered likely to have stabilized following the initial adaptation phase.
Clinical stability during this interval, defined as the absence of relapses, infections, or clinically significant physical deterioration, was also confirmed for all participants.
Measurement Error
Measurement error was quantified using the Standard Error of Measurement (SEM), calculated as SEM = SD × √(1−ICC), and the Smallest Detectable Change (SDC), calculated as SDC = 1.96 × √2 × SEM. Bland–Altman Limits of Agreement were computed as the mean difference between test and retest scores ± 1.96 × SD of the differences.
Construct Validity
Construct validity was evaluated by testing predefined hypotheses according to COSMIN recommendations.37 Structural validity, convergent validity, and known-groups validity were examined.
Structural Validity
The dimensional structure of both questionnaires was explored using Principal Component Analysis (PCA) with varimax rotation and Kaiser normalization. Sampling adequacy was assessed using the Kaiser–Meyer–Olkin (KMO) measure and Bartlett’s test of sphericity. The number of retained components was determined based on eigenvalues greater than 1 and inspection of the scree plot.
Convergent Validity
Since no gold standard exists for measuring exercise self-efficacy, convergent validity was examined by assessing the correlation between total EXSE-Sp (McAuley) and ESE-Sp (Neupert) scores using Spearman’s correlation coefficient (r_s), with 95% confidence intervals. A moderate to strong positive correlation (r_s ≥ 0.50) was hypothesized.
Known-Groups Validity
Known-groups validity was evaluated by comparing EXSE-Sp (McAuley) and ESE-Sp (Neupert) scores across three disability groups, defined according to EDSS classification (0–3.5, 4.0–6.5, and 7.0–8.5). The assumptions for between-group comparisons were assessed using the Shapiro–Wilk test, Q–Q plots, boxplots, and Levene’s test for homogeneity of variances. Given the unequal group sizes, the statistical approach was selected according to the distributional and variance assumptions for each outcome.
For the ESE-Sp (Neupert), one-way analysis of variance (ANOVA) was used, with Tukey’s post-hoc test for pairwise comparisons. For the EXSE-Sp (McAuley), Welch’s ANOVA was used because the assumption of homogeneity of variances was not met, with the Kruskal–Wallis test additionally performed as a sensitivity analysis. Based on previous literature suggesting that exercise self-efficacy may not be strongly determined by physical impairment alone,38 no substantial differences in self-efficacy scores were expected across disability groups.
Ethical Considerations
The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Ethics Committee of Hospital Universitari Vall d’Hebron, Barcelona, Spain (PR [AG] 419/2022). All participants provided written informed consent before enrollment.
Results
Descriptive Analysis
The sample had a mean age of 51.03 ± 9.69 years and comprised 60.2% women. It was characterized by long-standing disease (mean duration 17.1 years), a predominance of progressive MS phenotypes (61.3%), and moderate-to-severe disability (EDSS ≥4.0 in 76.3% of participants), which may limit the generalizability of the findings to the broader Spanish-speaking population of people with MS. Their demographic and clinical characteristics are presented in Table 1.Table 1Demographic and Clinical Characteristics of the Study Sample (N = 93)Age, Years, Mean (SD)51.03 (9.69)Gender, n (%)● female56 (60.2)● male37 (39.8)Education, years, mean (SD)13.45 (3.36)Marital status, n (%)● single14 (15.1)● married63 (67.7)● divorced15 (16.1)● widowed1 (1.1)Employment status, n (%)● student1 (1.1)● employed12 (12.9)● unemployed3 (3.3)● sick leave17 (18.3)● pensioner60 (64.5)Disease duration, years, mean (SD)17.10 (10.71)MS subtype, n (%)● Relapsing-Remitting36 (38.7)● Secondary Progressive40 (43.0)● Primary Progressive17 (18.3)EDSS (median)5.0● 0–3.5 n (%)22 (23.7)● 4.0–6.0 n (%)49 (52.7)● 6.5–8.5 n (%)22 (23.7)Abbreviations: SD, standard deviation; MS, multiple sclerosis; EDSS, Expanded Disability Status Scale.
Floor and Ceiling Effects
Both the EXSE-Sp (McAuley) and the ESE-Sp (Neupert) test and retest scores exhibited a negative skew. The EXSE-Sp (McAuley) showed a minor floor effect, with 7.5% and 8.6% of participants scoring the minimum possible score at test and retest, respectively, and a moderate ceiling effect, with 19.4% and 15.1% of participants scoring the maximum possible score at test and retest, respectively. In contrast, the ESE-Sp (Neupert) showed neither floor nor ceiling effects, with only 4.3% of participants scoring the minimum score at both test and retest, and 1.1% and 2.2% scoring the maximum score at test and retest, respectively (Table 2).Table 2Measurement Properties of the EXSE-Sp (McAuley) and ESE-Sp (Neupert) ScalesSectionIndicatorEXSE-Sp (McAuley) Score: 0–100ESE-Sp (Neupert) Score: 9–36Descriptive statisticsMean (SD), Test/Retest64.48 (31.68) / 63.30 (32.00)21.71 (6.83) / 21.38 (6.81)Median (IQR), Test/Retest72.5 (50–90)21 (16.5–26.5)Floor and ceiling effectsSkewness, Test/Retest−0.797 / −0.754−0.230 / −0.240Floor (%), Test/Retest7.5 / 8.64.3 / 4.3Ceiling (%), Test/Retest19.4 / 15.11.1 / 2.2ReliabilityInternal consistencyCronbach’s α (95% CI)0.977 (0.969–0.983)0.896 (0.861–0.925)Test–retest reliabilityICC (95% CI)0.992 (0.987–0.995)0.969 (0.954–0.980)Measurement errorSEM2.831.20SDC7.863.33Test–retest agreementMean difference (bias)1.180.33p-value (bias)0.0040.05795% CI of bias0.38 to 1.98−0.01 to 0.68SD of the differences3.891.67LLoA−6.45−2.95ULoA8.813.61Proportional bias (regression p-value)0.5690.402Construct validityStructural validityKMO0.8900.885Variance explained (%)86.6855.49Convergent validitySpearman’s rs, EXSE-Sp vs ESE-Sp0.56 (p < 0.001)Known-groups validityEDSS 0–3.5, Mean (SD)64.60 (31.89)20.50 (6.75)EDSS 4.0–6.5, Mean (SD)66.90 (28.18)22.30 (6.64)EDSS 7.0–8.5, Mean (SD)55.25 (42.83)21.27 (7.81)Primary statistical test, p-valueWelch’s ANOVA, p = 0.253One-way ANOVA, p = 0.637Sensitivity analysis, p-valueKruskal–Wallis, p = 0.437Kruskal–Wallis, p = 0.699Abbreviations: CI, confidence interval; EDSS, Expanded Disability Status Scale; ICC, intraclass correlation coefficient; IQR, interquartile range; KMO, Kaiser–Meyer–Olkin measure of sampling adequacy; LLoA, lower limit of agreement; SD, standard deviation; SDC, smallest detectable change; SEM, standard error of measurement; ULoA, upper limit of agreement.
Reliability
Internal Consistency Reliability
The EXSE-Sp (McAuley) demonstrated excellent internal consistency (Cronbach’s α = 0.977, 95% CI: 0.969–0.983), whereas the ESE-Sp (Neupert) showed good internal consistency (Cronbach’s α = 0.896, 95% CI: 0.861–0.925). Although the Cronbach’s alpha of the EXSE-Sp (McAuley) exceeded the commonly recommended upper threshold, this may reflect a high degree of item homogeneity (Table 2).
Item-total correlations and alpha-if-item-deleted analyses (Supplementary Material 3) showed that Item 9 of the ESE-Sp (Neupert) had a markedly lower corrected item-total correlation (r = 0.261) than the remaining items (0.562–0.751), and its removal would result in the smallest reduction in alpha, indicating limited contribution to the total score. Notably, this finding converges with the exploratory Rasch analysis, in which the same item showed misfit. Similarly, Items 1 and 2 of the EXSE-Sp (McAuley), which also showed misfit in the Rasch analysis, presented comparatively lower item-total correlations (0.781 and 0.816, respectively) than the remaining items (see Supplementary Material 2).
Test–Retest Reliability
Both questionnaires demonstrated excellent test–retest reliability, with ICCs of 0.992 (95% CI: 0.987–0.995) for the EXSE-Sp (McAuley) and 0.969 (95% CI: 0.954–0.980) for the ESE-Sp (Neupert). Bland–Altman analysis showed a mean difference (bias) of 1.18 points (95% CI: 0.38 to 1.98; p = 0.004) between test and retest scores for the EXSE-Sp (McAuley), indicating a small but statistically significant systematic difference, with 95% limits of agreement ranging from −6.45 to 8.81. For the ESE-Sp (Neupert), the mean difference was 0.33 points (95% CI: −0.01 to 0.68; p = 0.057), with 95% limits of agreement ranging from −2.95 to 3.61, indicating no statistically significant systematic bias.
Linear regression of the differences against the mean scores showed no evidence of proportional bias for either instrument (EXSE-Sp (McAuley): β = −0.007, p = 0.569; ESE-Sp (Neupert): R2 = 0.008, F(1.91) = 0.709, p = 0.402). Most observations fell within the limits of agreement for both instruments (Figure 1a, b and Table 2). Figure 1Principal component analysis of the EXSE-Sp (McAuley) and ESE-Sp (Neupert).
(a) Scree plot for the EXSE-Sp (McAuley). (b) Scree plot for the ESE-Sp (Neupert).Two scatter plots of difference versus mean for EXSE-Sp McAuley and ESE-Sp Neupert, 0 to 100.
Measurement Error
Both instruments showed low measurement error, reflected by a SEM of 2.83 points and an individual-level SDC of 7.86 points for the EXSE-Sp (McAuley), and a SEM of 1.20 points and an SDC of 3.33 points for the ESE-Sp (Neupert) (Table 2).
Construct Validity
Structural Validity
Principal Component Analysis (PCA) demonstrated adequate sampling adequacy, with KMO values of 0.890 for the EXSE-Sp (McAuley) and 0.885 for the ESE-Sp (Neupert). Bartlett’s test of sphericity was statistically significant for both questionnaires (p < 0.001). PCA identified a one-component solution for both questionnaires, based on the criterion of eigenvalues greater than 1.
Component loadings ranged from 0.769 to 0.975 for the EXSE-Sp (McAuley) and from 0.440 to 0.876 for the ESE-Sp (Neupert). The component extracted for each scale explained 86.68% and 55.49% of the total variance for the EXSE-Sp (McAuley) and ESE-Sp (Neupert), respectively (Figure 2). Figure 2Bland–Altman plots of test–retest agreement for the EXSE-Sp (McAuley) and ESE-Sp (Neupert).
(a) Bland–Altman plot for the EXSE-Sp (McAuley). (b) Bland–Altman plot for the ESE-Sp (Neupert). The red line represents the mean difference between test and retest scores; green lines represent the 95% limits of agreement (mean difference ± 1.96 sd).Two line graphs showing scree plots for EXSE Sp McAuley and ESE Sp Neupert.
Convergent Validity
A moderate and statistically significant correlation (r = 0.56, p < 0.001) was observed between the EXSE-Sp (McAuley) and the ESE-Sp (Neupert), supporting convergent validity (Table 2).
Known-Groups Validity
Known-groups validity analysis showed no statistically significant differences in both questionnaires’ scores across disability levels (EDSS 0–3.5, 4.0–6.5, and 7.0–8.5). For the ESE-Sp (Neupert), homogeneity of variances was met (Levene’s test, p = 0.117), and the one-way ANOVA showed no significant differences between EDSS groups, F(2.90) = 0.452, p = 0.637. Tukey’s post-hoc comparisons showed no significant differences between any pair of groups (all p > 0.05).
For the EXSE-Sp (McAuley), homogeneity of variances was not met (Levene’s test, p = 0.003). Therefore, Welch’s ANOVA was used and showed no significant differences between EDSS groups, F(2.380.753) = 1.424, p = 0.253. The Kruskal–Wallis sensitivity analysis yielded a consistent result, χ2(2) = 1.655, p = 0.437 (Figure 3 and Table 2).
Figure 3Distribution of exercise self-efficacy scores across disability groups (EDSS 0–3.5, 4.0–6.5, and 7.0–8.5) for the EXSE-Sp (McAuley) and ESE-Sp (Neupert).Box plots showing exercise self-efficacy scores across EDSS score groups for two questionnaires.
Discussion
The present study provides evidence supporting the measurement properties of the Spanish versions of the Exercise Self-Efficacy Scale by McAuley (EXSE-Sp) and the Exercise Self-Efficacy Scale by Neupert (ESE-Sp) in Spanish PwMS. By addressing the lack of culturally adapted and validated instruments for assessing exercise-related self-efficacy in Spanish neurological populations, this study contributes to the availability of patient-reported outcome measures (PROMs) that can facilitate both clinical decision-making and research on exercise participation in PwMS.
Principal component analysis supported a unidimensional structure for both instruments, consistent with the theoretical conceptualization of exercise self-efficacy as a relatively specific and coherent construct. Although the EXSE-Sp (McAuley) explained a higher proportion of variance compared with the ESE-Sp (Neupert) (86.68% vs 55.49%), both scales demonstrated adequate item loadings and acceptable structural validity. The lower explained variance observed for the ESE-Sp (Neupert) may reflect the greater heterogeneity of its items, which assess confidence in exercising across a range of distinct barriers and circumstances, whereas the EXSE-Sp (McAuley) items refer more specifically to maintaining a single, defined exercise behaviour.
This broader item content in the ESE-Sp (Neupert), while resulting in comparatively lower explained variance, may also allow it to capture a wider range of self-efficacy-related experiences. These findings are consistent with previous studies suggesting that exercise self-efficacy can be appropriately represented as a unidimensional construct in individuals with chronic conditions, including neurological populations such as PwMS.39,40
Both instruments demonstrated high internal consistency, with Cronbach’s alpha values of 0.977 for the EXSE-Sp (McAuley) and 0.896 for the ESE-Sp (Neupert). While values above 0.80 are generally considered indicative of good internal consistency, coefficients above 0.90 may suggest a degree of item redundancy rather than improved measurement precision.41 This issue has been previously discussed in the context of PROM development, as highly homogeneous items may provide overlapping information and artificially increase alpha coefficients.42 The COSMIN framework also recommends interpreting very high internal consistency values alongside other indicators of structural validity and measurement performance.43 In this sense, the high internal consistency observed for the EXSE-Sp (McAuley) is consistent with findings reported in similar populations with chronic conditions,44,45 but it also raises the possibility that certain items may be redundant and could potentially be removed without compromising conceptual coverage. Nevertheless, we retained the original item structure to preserve comparability with previous studies and to ensure comprehensive assessment of the construct.
Furthermore, both instruments showed low measurement error, as reflected by their small SEM and SDC values. These results suggest that changes exceeding the SDC are likely to represent true changes beyond measurement error, supporting the use of both instruments for longitudinal assessment in both clinical practice and research. Future studies may explore whether shorter versions could maintain adequate measurement properties while reducing respondent burden.
A moderate ceiling effect was identified for the EXSE-Sp (McAuley), which may reduce its ability to discriminate between individuals with high levels of exercise self-efficacy. This finding could be related to the nature of the questionnaire, which assesses confidence in maintaining a specific exercise routine, potentially resulting in higher scores among individuals who have recently completed rehabilitation or have increased familiarity with structured exercise.22 Notably, this may also reflect a characteristic of the sample and timing of assessment: the EXSE-Sp (McAuley) and ESE-Sp (Neupert) were administered near the end of participants’ rehabilitation program, shortly before discharge, specifically to capture their confidence in continuing to exercise independently once supervised rehabilitation ended. Having just completed a structured exercise program, participants may have felt particularly confident about their exercise capabilities at this specific point in time, which may have contributed to the ceiling effect observed for the EXSE-Sp (McAuley).
Whether this elevated self-efficacy is maintained once patients transition out of the supervised rehabilitation setting remains to be established. Conversely, the absence of relevant floor or ceiling effects in the ESE-Sp (Neupert) suggests that this instrument may capture a wider range of self-efficacy perceptions, particularly in relation to overcoming exercise barriers, and may therefore be less sensitive to this sample- and timing-related ceiling. This distinction is consistent with the conceptual focus of each instrument: the EXSE-Sp (McAuley) primarily assesses confidence in maintaining exercise behavior over time,22 whereas the ESE-Sp (Neupert) evaluates confidence in exercising despite potential barriers or challenging circumstances,23 which may make it comparatively less susceptible to ceiling effects driven by a single, recently completed rehabilitation experience.
The moderate correlation observed between the EXSE-Sp (McAuley) and ESE-Sp (Neupert) supports their convergent validity while also indicating that both instruments assess complementary dimensions of exercise self-efficacy. Therefore, the selection of one instrument over the other may depend on the specific clinical or research objective.46,47 The EXSE-Sp (McAuley) may be particularly useful for monitoring changes during structured exercise programs, whereas the ESE-Sp (Neupert) may provide additional information regarding perceived barriers and behavioral regulation in everyday contexts.
The absence of significant differences in self-efficacy scores according to disability severity provides additional insight into the multidimensional nature of exercise participation in PwMS. These findings suggest that exercise self-efficacy may not be directly determined by physical impairment alone. Instead, psychological and contextual factors, including motivation, previous exercise experiences, perceived barriers, and social support, may play an important role in shaping individuals’ confidence to engage in exercise.36 This reinforces the importance of assessing self-efficacy independently from disability status when designing patient-centred interventions.38
The availability of validated Spanish versions of the EXSE (McAuley) and ESE (Neupert) represents an important step toward improving the assessment of psychosocial determinants of exercise behavior in the Spanish population of PwMS. Incorporating these patient-reported measures into routine assessment promotes a more patient-centered approach by acknowledging individuals’ perceptions, confidence, and personal barriers as relevant components of exercise participation and rehabilitation planning.48 Self-efficacy is a modifiable factor that may influence exercise initiation, adherence, and long-term participation.13,22,29 Thus, these instruments may help identify individuals requiring additional support and inform personalized strategies to strengthen confidence and address exercise-related barriers. This is particularly relevant in PwMS, for whom maintaining long-term engagement in PA remains challenging despite the recognized benefits of exercise for symptom management, functional outcomes, and quality of life.49 More broadly, the use of culturally adapted PROMs may support the implementation and evaluation of effective and sustainable exercise interventions.49
Several limitations should be acknowledged. First, the exploratory Rasch analysis was limited to item-level fit and point-measure correlations and did not include formal assessments of dimensionality, local dependency, differential item functioning, or category/threshold ordering.
Second, the present study assessed the measurement properties of both instruments but did not examine their association with objectively or subjectively measured PA. Therefore, the clinical relevance of exercise self-efficacy as a determinant of actual exercise participation in this population remains theoretical and should be addressed in future research.
Third, this study did not assess responsiveness. Given that these instruments are intended to support the evaluation of interventions aimed at promoting long-term PA participation, this represents an important limitation that constrains the applicability of the present conclusions until responsiveness has been established.
Fourth, participants were recruited from a single specialist rehabilitation centre, and the questionnaires were administered shortly before discharge, specifically to assess confidence in maintaining exercise once supervised rehabilitation had ended. Compared with community-based PwMS, this sample may therefore differ in terms of disability, rehabilitation exposure, and timing of assessment. In addition, the timing of the assessment may have contributed to an elevation in self-efficacy following recent completion of a structured exercise program.
These characteristics may have influenced the reliability estimates, the ceiling effect observed for the EXSE-Sp (McAuley), and the construct validity correlations, and may limit the generalizability of the findings to the broader Spanish MS population.
Fifth, the test–retest interval was scheduled during the final two weeks of the participants’ rehabilitation program. Although clinical stability was confirmed, participants remained in active rehabilitation during this period, and the assumption that exercise self-efficacy had stabilized was not empirically verified. Therefore, the possibility that construct-level changes related to the rehabilitation process occurred cannot be completely excluded.
Sixth, a small but statistically significant systematic bias was identified between test and retest scores for the EXSE-Sp (McAuley), which should be considered when interpreting individual-level change over time.
Finally, self-efficacy is influenced by contextual and individual factors that may change over time, such as previous exercise experiences, environmental barriers, and social support.
Conclusions
This study provides evidence supporting the reliability and validity of the Spanish versions of the Exercise Self-Efficacy Scale by McAuley (EXSE-Sp) and the Exercise Self-Efficacy by Neupert (ESE-Sp) in a sample of PwMS recruited from a specialist rehabilitation centre in Spain. Both instruments demonstrated adequate measurement properties and represent potentially useful tools for assessing exercise-related self-efficacy in Spanish clinical populations. However, further validation in community-based samples and across other Spanish-speaking countries is needed before these findings can be generalized more broadly.
Their availability may facilitate the identification of individuals with reduced confidence in exercise participation and support personalized intervention planning. Their use in evaluating interventions and promoting long-term PA participation will require future studies establishing their responsiveness. By providing culturally adapted PROMs, this study represents an initial step toward advancing evidence-based approaches to assessing psychosocial determinants of PA participation among Spanish PwMS.