The study was approved by the Reginal Committee for Medical and Health Research Ethics, Norway (REK Nord 174837) and the Nordland Hospital Research Ethical Authority (project number 209). We conducted in accordance with the Helsinki declaration, and all participants provided written informed consent prior to inclusion. The study was funded by the Nortern Norway Regional Health Authority (Grant number 174837). The funder played no role in the design, conduct or reporting of the study.

The adaptation of the MSWDQ-23 (19) into Norwegian, followed the recommendations made by the International Society for Pharmacoeconomics and Outcomes Research (33), containing nine steps, described in the following.

(1) First, we obtained permission from the developer of MSWDQ-23 who agreed to participate as a consultant. Then we established a team consisting of one key person (ECA) who worked closely with the project manager (BN) throughout the adaptation process, two native Norwegian physical therapists to conduct the forward translation of whom one had studied in English speaking countries. For the back translation, we recruited one other native Norwegian physical therapist, who had studied in English speaking countries, and one native English speaker who was also fluent in Norwegian and worked as a professional translator. (2) The team members independently translated the questionnaire into Norwegian and made comments regarding ambiguities or word choice. (3) Reconciliation: The team members discussed the translated drafts and recognized some variations and challenges. The Norwegian version introduced some language changes in the instructions and some of the items. In Norwegian the word “please” is not so commonly used while giving instructions. We decided, in line with the Spanish version, to preserve the word. However, we did not repeat it twice in the last sentence of the introduction. When translating “I found” and “I thought” we used the same Norwegian word (“jeg syntes”). We used job (“jobb”) instead of work and leader (“leder”) instead of manager. Moreover, we discussed the differences between “I thought”, “I felt”, “I experienced”, “I found”, “I feared” with the original MSWDQ author (CH) to ensure that the translation covered the intended meaning. Since the Norwegian health care system implies that there is no immediate loss of payment, we, in agreement with CH, changed item No.21 (corresponding No.48 in the original 50-item scale) to include future decrease of payment. (4) By involving both a professional with knowledge in MS and experience of treating these patients, as well as a native English person with professional linguistic knowledge in the back translation of the reconciled translation of the questionnaire, we aimed for both conceptual, cultural as well as linguistic equivalence. (5) Back translation review was first discussed within the team and then with the developer of the original MSWDQ-23 (CH). (6) Harmonization was conducted by the project manager (BN) and the key person (ECA) by comparing the original, the Norwegian version and the back translated version. (7) Cognitive debriefing was obtained by distributing the Norwegian version of MSWDQ-23 to a group of pwMS (n = 6) who completed the questionnaire and were telephone interviewed about their item interpretation/understanding. (8) The whole team reviewed the cognitive debriefing results and made some adjustments. (9) Proofreading was performed by the BN and ECA. An overview of the process is shown in Figure 1. The MSWDQ-23 Norwegian Version (MSWDQ-23NV) is available as Supplementary Materials.

No floor effects with the best possible score of zero on the MSWDQ-23 were present (5.2% of participants had scores of 0 and no participant had scores of 100). For the individual subscales, while no floor effects were present with the psychological/cognitive and physical barriers subscales (8.7% and 8.3%, respectively with best possible scores) a floor effect was present in the external barriers subscale, with 15.3% of participants indicating the best possible score of zero. There were no ceiling effects present in the total MSWDQ-23 or subscales.

An initial CFA produced the following statistics: χ²/df = 3.016, RMSEA = .098 (lower bound = .089, upper bound = .106), SRMR = 0.0736, CFI = 0.809, and BIC = 846.96, indicating a poor fit of the data. Following inspection of the modification indices, several conceptually linked error terms were correlated (19, 22). In the psychological/cognitive barriers subscale these included: “forgot what task”, “needed to be reminded” and “remember conversation”; and “employer not understanding” and “colleagues not supportive”. In the physical barriers subscale these included: “disturbances in bowel/bladder” and “feared incontinent”. In the external barriers subscale these included: “feared no longer support self” and “reduce work hours, reduced pay”. Correlating the error terms improved the model fit as follows: χ²/df = 1.979, RMSEA = .068 (lower bound = .059, upper bound = .077), SRMR = 0.0692, CFI = 0.910, and BIC = 613.347. These indices represent overall fair model fit for the data, thus supporting the structure of the MSWDQ-23. Modification indices pertaining to item loadings were also inspected to see if any removal would improve fit indices, particularly the lower bound of the RMSEA which remained slightly outside the requirements for good fit. The removal of items “perform to level expected” and “became sleepy” in the psychological/cognitive barriers subscale improved the fit further as follows: χ²/df = 1.796, RMSEA = .061 (lower bound = .050, upper bound = .072), SRMR = 0.0686, CFI = 0.931, and BIC = 483.957. However, given the very high correlation (r = .988) between the scores for the psychological/cognitive barriers subscale with and without these items, this improvement was determined to be negligible, hence supporting the inclusion of all 23 items in the scale. Table 2 shows factor loadings for the MSWDQ-23.

Correlations between the three subscales were large (psychological/cognitive barriers and physical barriers, r = .611; psychological/cognitive barriers and external barriers, r = .846; and physical barriers and external barriers, r = .596). Given the very large correlation between the psychological/cognitive barriers and external barriers, an alternative model with respective items for these subscales loading onto a single factor was conducted in CFA. However, this did not improve the fit of the model [χ²/df = 2.119, RMSEA = .073 (lower bound = .064, upper bound = .082), SRMR = 0.0692, CFI = 0.896, and BIC = 643.955].

A second-order CFA with all 23 items loading onto the three factors, and in turn, a further higher-order overall MSWDQ factor was also examined. Fit statistics for this model were identical to the first order model and were as follows: χ²/df = 1.979, RMSEA = .068 (lower bound = .059, upper bound = .077), SRMR = 0.0692, CFI = 0.910, and BIC = 613.347.

Cronbach's α values for the MSWDQ-23 subscales are shown in Table 2. All subscales had excellent internal consistency. No removal of any item would substantially improve these internal consistency values. Cronbach's α was.94 for the full MSWDQ-23.

The Norwegian MSWDQ-23 physical, psychological/cognitive, and external barriers subscales, had large, small-to-moderate, and small positive relationships respectively with EDSS scores (46). Total MSWDQ-23 scores were moderately and positively related to EDSS. There were large relationships present between the MSWDQ-23 total and subscales and the physical and the mental health composite scores. There were moderate-to-large relationships present between Total MSWDQ-23 and psychological/cognitive barriers scores and employment status. Physical barriers and external barriers had a moderate and small relationship respectively with work status (%). Table 3 shows correlations with the alternative scales used in this study.

MSWDQ total and subscale scores according to MS type, together with inferential statistical information, is shown in Table 4. While differences in MSWDQ-23 scores were not detected in the psychological/cognitive barriers and external barriers subscales, and the full MSWDQ-23, group differences were found in the physical barriers subscale. In particular, the participants with RRMS had significantly fewer physical work difficulties than both the SPMS and PPMS participants. There were no differences between SPMS and PPMS in physical barriers scores. MSWDQ total and subscale scores according to employment status, together with inferential statistic information, is shown in Table 5. The participants who were working over the previous 12 months had significantly fewer work difficulties than people who had not been working over the previous 12 months.

A clear limitation of the present study is that we did not examine test–retest reliability, which is due to the selected cross-sectional design. However, prior studies have reported a high test–retest reliability for the MSWDQ-23 (21, 23, 28). Thus, there are indications of likely adequate test–retest reliability for the Norwegian version of the MSWDQ-23 but future studies should examine this. Moreover, we did not include healthy controls in our study, which would have helped us attribute the reported work difficulties to MS. Although one of the available MSWDQ-23 validation studies showed good discrimination (27), and notwithstanding the fact that the scale asks participants to report on difficulties that are attributable to their MS, the hypothesis that the reported difficulties in the Norwegian study extend beyond those experienced by healthy individuals was not tested. Finally, this study did not investigate the future employment status of the participants, which would have allowed us to confirm the predictive validity of this instrument. However, previous studies (19, 22) have demonstrated the predictive validity of the instrument for future employment.

A strength of the current study is that we used a confirmatory factor analysis technique to verify the three-factor model structure and evaluated it against both Kline (47) and Hu and Benter (41), documenting a fair fit. Moreover, an advantage of the current study was the inclusion of persons with all types of MS, low EDSS (2.81), and both employed and not employed pwMS. Additionally, the relatively large sample (n = 229) can be considered a strength given it is likely to be representative of the Norwegian MS population.