Table 1 Summary of psychometric properties evaluated and the criteria used for determining the adequacy of the MSRMS

Item analysis and scale development

Items were eliminated due to:

high missing data (> 10%);

maximum endorsement frequencies (percentage of responses for the most frequently endorsed category) > 80%;

sum of endorsement frequencies for any 2 adjacent item response categories < 10%;

For each item that correlated highly with another item (> = 0.75), the item with the least favourable

psychometric

properties (on the other criteria above) was eliminated

In principal components factor analysis (a method for reducing data while retaining those characteristics of the dataset that contribute most to its variance):

Items that did not correlate (< 0.30) with the first unrotated component were removed, as it indicates that the items

are not measuring a single underlying construct

Items that did not load on any factors (< 0.40) or cross-loaded on two or more factors (< 0.20) were removed (the aim was

to achieve a simple structure, that is, each item should be associated with one and only one factor)

Data quality

Missing item data < 10%

High % computable scale scores

Scaling assumptions

Similar response-option frequency distributions

Similar mean scores and variances

Similar and substantial (r > 0.30) item-total correlations [14]

Targeting

Scores span the full scale range [15]

Mean scores near the midpoint [16]

Floor and ceiling effects (endorsement at the bottom and the highest end of the scale) < 20% [17]

Reliability

Cronbach's alpha (an indicator of internal consistency, or the extent to which items are interrelated) > 0.70 [18]

Validity

Discriminant construct validity, (evidence that the scale do not correlate with dissimilar measures) evidenced by low correlations (r < 0.30) between all subscales of the MSRMS and the Multiple Sclerosis Impact Scale (MSIS-29) [9] scales and the Medical Outcomes Study Short-form Health Survey SF-36 scales (SF-36) [19]

Group differences validity, (the ability of the MSRMS scores to differentiate groups expected to differ) evidenced by stepwise mean differences in MSRMS scores for groups defined by patients' response to the single item question: "Overall, how would you rate the quality of care you received for your relapse? 1 = poor, 2 = fair, 3 = good, 4 = very good, 5 = excellent"