Psychometrics of the Wrist Stability and Hand Mobility Subscales of the Fugl-Meyer Assessment in Moderately Impaired Stroke

Discussion

The genesis of promising regimens targeting the paretic UE necessitates scales to measure efficacy of such approaches. Some stroke measures^1,2 do not capture the motor gains derived from these treatments, and others are validated in people with active distal UE movement^7–9—an ability exhibited by only a minority of survivors of stroke.^22,23 The current study examined w/h UE FM psychometrics in patients with only palpable wrist flexion, hoping to extend the number of patients with stroke who could be rigorously tested with a bedside measure. This was the first study assessing the w/h UE FM's psychometric characteristics in this segment of the rapidly expanding stroke survivor population.

Results suggest high w/h UE FM intrarater reliability and item consistency in this group. Because there was considerable time and many other test administrations for this and other studies between a particular participant's first and second UE FM administrations, the possibility of memory bias (ie, the rater recalling a particular participant's first rating and allowing it to influence his or her second rating) was low. Additionally, the reliability and concurrent validity values obtained in the current study compare favorably with UE FM and ARAT reliability and validity values in previous work,^14,24 although individuals in previous studies exhibited more UE movement than those in the current study. Thus, our study is a “first” in the literature not only for its analyses of the w/h UE FM but also in that it is the first to examine ARAT psychometrics in patients with only palpable wrist flexor activity.

Although psychometric values met the criteria that were set forth a priori, the second pretesting session's concurrent validity values were slightly less than the .70 criterion (P=.67). However, regression-adjusted Bland-Altman plots, used to graphically explore agreement, revealed that few observations (1 and 2 of 32 observations, respectively, between pretest measurements) resided outside the 95% limits of agreement, with these points lying close to the boundary of the limits. Furthermore, points in the Bland-Altman plots were scattered both above and below the mean value, indicating that bias was not overwhelmingly in one direction or the other. These findings confirmed that the w/h UE FM exhibited adequate concurrent validity with the ARAT and further support its use as a “stand-alone” bedside scale to measure distal paretic UE movement.

Cronbach alpha is an established²⁵ measure of internal consistency. It is commonly used to assess reliability during the development of new measures.²⁶ However, recent investigations^19,20,27 demonstrated that the alpha coefficient may be underestimated when fewer than 5 scale points are used. This finding is particularly concerning within rehabilitation contexts, where Likert-type response scales are often used to measure function and clinical outcomes (eg, FM, ARAT). Specifically, Cronbach alpha assumes that item responses will be normally distributed and measured using interval or ratio scales. This assumption preents 2 problems. First, item responses in clinical settings may not be normally distributed. For example, the distributions of responses of survivors of severe stroke to an assessment of motor impairment are unlikely to be normally distributed. Rather, such responses are likely to be positively skewed, with the majority of cases indicating significant impairment of motor function. Distributional analyses of the FM data (pretest 1, pretest 2) collected as part of this study confirm this hypothesis. Second, Cronbach alpha is inappropriate because Likert-type responses are never continuous.²⁸ The violation of this assumption is confirmed based on visual inspection of distributional analyses.

Our use of reliability analyses of the w/h UE FM using ordinal alpha are significant for a number of reasons. First, the estimates of ordinal alpha exceed those of Cronbach alpha (Table). These findings confirm the expectation presented in previous work on ordinal alpha.^18,19 Second, to our knowledge, this article presents the first application of ordinal alpha within the field of physical rehabilitation. This reason is significant because of the high prevalence of Likert-type measurements used to measure health, function, and clinical outcomes in the rehabilitation professions. Based on our findings, the use of ordinal alpha is highly recommended for therapists, physicians, and other rehabilitation scientists conducting clinical research that uses Likert-type response scales. This approach allows researchers to account for the discrete nature of data generated by these types of instruments. Third, because Cronbach alpha may sometimes underestimate the internal consistency of an instrument (although it is not believed to have done so in the current study), clinicians may inaccurately determine that a measure is unreliable when it actually is reliable. Finally, analyses conducted using ordinal alpha further confirm the high reliability of the w/h UE FM.

When coupled with the administration instructions provided elsewhere,¹⁰ it is hoped that the above findings will encourage clinicians to use the w/h UE FM to quantify UE movement in patients exhibiting at least palpable paretic wrist flexor activity. Clinicians can use w/h UE FM scores to rigorously assess current UE movement, to plan treatments, and to gauge efficacy of an intervention by administering the test longitudinally. For example, deficiencies seen in items measuring mass flexion and mass extension may give the clinician reason to target grasping and releasing a coffee cup as part of therapy sessions. Moreover, motor performance constitutes a significant predictor of eventual motor recovery in the months poststroke.²⁹ As next steps, it would be prudent to examine predictive validity and responsiveness of the measure as well as its interrater reliability, as only one rater was used in this study. Examining sensitivity to change, such as when an intervention is provided, also would be a desirable next step.

Like other measures of distal UE movement,^8,9 the ARAT requires patients to be seated in a chair and uses a specific setup for patient and object positioning. Although important for research purposes, these conditions may not always be plausible in clinical situations. These shortfalls also make the w/h UE FM well situated as a quick screen or outcome measure in research situations. Yet, it should be noted that the w/h UE FM is a measure of UE impairment, whereas other measures (eg, ARAT, Arm Motor Ability Test) examine UE functional limitation. Although scores on these measures are highly associated, they do not measure entirely equivalent constructs (ie, motor impairment with the w/h UE FM versus functional limitation using the latter measures). Care should be taken by researchers to consider the construct in which they expect and hope to measure changes when using these measures. It should be noted that a short form of the UE FM, consisting of 12 items, was recently validated.³⁰ However, inclusion of items on this particular version were predicated on the premise of proximal to distal motor return, which has been shown to not always be valid.¹¹ Moreover, one must still administer the more proximal items on this and other revised UE FM versions³¹ that have been forwarded. The current study remains valuable because it examines the administration of the w/h UE FM as a stand-alone measure, without the more proximally focused items. Clinicians will want to choose the best UE FM version depending on the impairment level of the patient, the time, and the tools that they have available.