Our study shows that the SwePASS is responsive to change, as based on the results of assessments during the first week and 3 months after stroke, at least for those patients who do not reach a maximum score for the test at baseline. Moreover, the results showed that the patients improved in terms of postural control, as assessed using the SwePASS score, up to 6 months after stroke. We also found a slight deterioration in postural control, according to the total SwePASS score, between 6 and 12 months after stroke.
Although most of the patients improved in postural control during the first 3 months after the stroke, some patients deteriorated. The amount of training might differ, but it is unclear how much and how this influenced postural control assessed using SwePASS. The median age at inclusion in the study was 73 years, and several of the patients had other diseases and conditions that might have influenced postural control.
“Postural control” must be discussed to be able to relate the results to clinical settings. What does “postural control” mean and what has been estimated? We have adopted the definition of postural control of Shumway-Cock and Woollacott  as an integration of different sensory modalities for body position in space (stability) and different motor strategies for the choice of body movement (appropriate orientation of body segments), as well as attention processing, which constitutes the basis on which functional movement is achieved. In addition, postural control is described as emerging from an interaction of the individual with the task and the environment .
Our data emphasize the fact that trustworthy information about the change (a real change or not) in postural control in stroke patients can be captured using the SwePASS score. As the SwePASS score shows responsiveness to change, it is feasible for use in clinical settings. Previously published data [12, 13] support this statement. Since rating scales have a fundamental role in the determination of patient care and in the evaluation of the effects of clinical interventions, besides being able to serve as an expression of clinical professionalism, these findings are of importance.
Our results concerning responsiveness are in line with the results presented by Yu et al.  and Mao et al. . Yu et al.  found high internal responsiveness during the hospital stay in 85 patients severely disabled by stroke, with a median total PASS admission score of 16 (min 0-max 36). The time of the initial evaluation was a median of 19 days after stroke onset. Mao et al.  found a high responsiveness of PASS in 93 patients between 14 and 90 days after stroke. Compared with the mean SwePASS admission score of 28 (SD 7.1) in the present study, Mao’s population had greater stroke disabilities, with a mean total PASS admission score of 17.6 (SD 12.8). In addition, Mao et al.  had a slightly younger population with a higher percentage of cerebral haemorrhage (26%). Both Yu et al.  and Mao et al.  showed a low percentage of patients that received the maximal score (ceiling effect). Our study also showed a low ceiling effect of 14%, although there were considerably higher admission scores. However, even though we showed that the SwePASS score is responsive, our results in terms of the higher percentage of ceiling effects seen with the longer passage of time since stroke (28% at 3 months and 40% at 6 months) indicate that the SwePASS score has a reduced ability to discriminate postural control beyond 3 months after stroke. When selecting the SwePASS to estimate changes in postural control, it might not be the first choice for patients who are at the end of impairment, with mildly impaired postural control. A modification of item categories or a greater number of item categories might increase the ability of SwePASS to detect changes. This is just speculation, however.
The PASS was also shown by Benaim et al.  to be less suitable after 3 months, as based on the distribution of scores, where 38% of 58 patients had already received the maximal score 90 days after stroke.
Only item 8 and the total SwePASS score showed a significant systematic change in the positions of scores from 3 to 6 months. Our result regarding recovery does not completely correspond with that of Jørgensen et al.  who, based on a large population study, reported that recovery in Activities of Daily Living was completed within 3 months in the majority of the patients and that recovery in severely impaired patients should not been expected after 5 months. Another interesting finding was that 10 to 25% of the patients showed a deterioration in the total SwePASS score, with a higher percentage of deterioration occurring over a longer time since the onset of stroke; this result is in line with the known non-linear pattern of the function of time . One possible explanation for this, perhaps in combination with older age and a natural decline in function, might be that patients receive more intensive rehabilitation interventions at the stroke unit than after discharge, with ambiguity in terms of who is responsible for continued rehabilitation intervention after discharge.
This study has some limitations. First, some patients were lost to follow-up. Offering follow-up investigations at the patients’ home would probably have decreased the number of patients lost to follow-up. Second, it would have been interesting to have several close follow-ups during the early phase, since we do not know when (especially during the first 3 months after stroke) the major part of the improvement took place. Third, more than one physiotherapist was involved in the assessments. However, we believe that any impact on the results is limited, in view of the previously published high interrater reliability  of the SwePASS, besides clear instructions to the physiotherapists prior to their involvement in the assessments.
Finally, we cannot differentiate between functional improvements due to spontaneous neurological recovery and improvements following rehabilitation interventions. However, in a randomized controlled trial with an intervention of 80 hours of physiotherapy vs. self-initiated exercise during the first year following stroke, Langhammer et al.  found a steady improvement in balance (as well as in Instrumental Activities of Daily Living (IADL), motor function, gait parameters and grip strength) up to 6 months after stroke, regardless of the group assignment.
Possible baseline variability may have influenced responsiveness; however, the time of the initial assessment was chosen using the hypothesis that the patients should be clinically stable (no changes due to penumbra or cerebral oedema). However, motor recovery may already have taken place during the first few days after stroke, and thus some patients may have recovered before the first assessment. Further research with larger populations and closer follow-up assessments are desirable in order to verify our findings and describe the time course of improvements in postural balance. With a focus on the significance of the one-leg standing items, it would be very interesting to investigate whether or not the SwePASS could be reduced to fewer items and still be able to give the same information.
We employed Svensson’s method  since it was exclusively developed to make calculations based on ordinal data and to use all the data information. Although more classical methods (Mc Nemar’s test, sign test etc.) are applicable for ordinal data, they require dichotomization of the data, which means that information from the other categories is missed. In addition, Svensson’s method makes it possible to recognize systematic disagreement caused by individual variability in assessments separately from systematic disagreement related to the group. A disadvantage in using Svensson’s method might be that changes in RP can be difficult to interpret in a clinical setting. An alternative method would be to use Rasch analysis , which might add value as a method of examining and comparing rating scale responsiveness . Rasch analysis is a method which, if a scale is revealed to be a unidimensional scale, could provide a non-linear transformation of the scale’s ordinal raw score to interval measure. Interval data allow an accurate interpretation of change in scores and the use of parametric statistics. To improve the clinical interpretation of change in postural control, it would have been interesting to transform the values of systematic changes in position (RP) into raw SwePASS scores. This was not possible with the present data, however. One potential method could be to perform the Rasch analysis  on a large dataset. Finally, the responsiveness shown for the SwePASS, distribution based established, might be even more clinically meaningful if supported by person-centred responses, i.e. anchor based, using subjective experiences to answer the question of whether or not the change in postural control is important.