Are the Hierarchical Properties of the Fugl-Meyer Assessment Scale the Same in Acute Stroke and Chronic Stroke?
- J. Lesley Crow,
- Gert Kwakkel,
- Johannes B.J. Bussmann,
- Jos A.G. Goos and
- Barbara C. Harmeling-van der Wel;
- for the Early Prediction of Functional Outcome After Stroke (EPOS) Investigators
- J.L. Crow, MSc, MCSP, DipTP, Unit Physical Therapy, Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC University Medical Center Rotterdam, Room Nc-318, PO Box 2040, 3000 CA Rotterdam, the Netherlands.
- G. Kwakkel, PT, PhD, Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, the Netherlands, and Department of Neurorehabilitation, Reade Center for Rehabilitation and Rheumatology, Amsterdam, the Netherlands.
- J.B.J. Bussmann, PT, PhD, Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC University Medical Center Rotterdam.
- J.A.G. Goos, PT, MNR, Department of Physical Therapy, Franciscus Hospital, Roosendaal, the Netherlands.
- B.C. Harmeling-van der Wel, PT, Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC University Medical Center Rotterdam.
- Early Prediction of Functional Outcome After Stroke (EPOS) Investigators (see list of investigators in the Footnotes section).
- Address all correspondence to Ms Crow at: j.crow{at}erasmusmc.nl.
Abstract
Background The motor function section of the Fugl-Meyer assessment scale (FM motor scale) is a robust scale of motor ability in people after stroke, with high predictive validity for outcome. However, the FM motor scale is time-consuming. The hierarchical properties of the upper extremity (UE) and lower extremity (LE) sections of the FM motor scale have been established in people with chronic stroke. These data support the use of a more concise method of administration and confirm scores can be legitimately summed.
Objective The aim of this study was to establish that a similar hierarchy exists in people within 72 hours after stroke onset.
Design A prospective, cross-sectional design was used.
Methods Data were obtained from 75 eligible people in a nationwide prospective study (the Early Prediction of Functional Outcome After Stroke). The full version of both sections of the FM motor scale was administered within 72 hours after stroke onset. The hierarchy of item difficulty was investigated by applying Guttman scaling procedures within each stage and each subsection of the UE and LE sections of the scale. The scaling procedures then were applied to item difficulty between stages and subsections and finally across all scale items (stage divisions ignored) of the FM motor scale.
Results For all analyses, the results exceeded acceptable levels for the coefficient of reproducibility and the coefficient of scalability.
Limitations The sample was a population of people with stroke of moderate severity.
Conclusions The unidimensional hierarchy of the UE and LE sections of the FM motor scale (already established for chronic stroke) within 72 hours after stroke onset was confirmed. A legitimate total summed score can indicate a person's level of motor ability.
Evidence-based practice requires accurate and valid tools to assess a patient's status and prognosis and to evaluate treatment. In patients with motor dysfunction after stroke, a reliable assessment of motor ability can assist in planning of treatment and monitoring of progress. The fact that changes in motor ability at the level of impairment can be recorded up to 7 days before functional gains1 provides a useful assessment level for the early detection of changes. Furthermore, impairment-level outcome measures are assumed to be the measures most closely related to the volume of brain loss and the best markers of prognosis after stroke.2 However, the use of measures in clinical practice is determined not only by methodological characteristics but also by practical issues, such as quick and easy administration.
The upper extremity (UE) and lower extremity (LE) sections (excluding balance) of the motor function domain of the Fugl-Meyer assessment scale (FM motor scale)3 are primarily used to assess the impairment of motor ability. The FM motor scale score is seen as a reflection of true neurological repair after stroke, mainly indicating spontaneous neurological recovery after stroke.4–6 Therefore, the FM motor scale is used as a prognostic indicator for the return of motor activity.7,8 The underlying conceptual framework of the FM motor scale is a stepwise order of motor recovery based on sequential stages, as described by Twitchell9 and Brunnström.10 In each section, scale items are ranked according to increasing item difficulty by means of stages and subsections, which form a stepwise progression. For the UE, the shoulder-elbow-forearm subsection is divided into 5 stages that follow the order of returning activity. This process begins with reflex activity and then follows the influence of flexor and extensor synergies to normal volitional movement. The remaining subsections are wrist, hand, and coordination. The LE has just 2 subsections, the hip-knee-ankle subsection and coordination; the hip-knee-ankle subsection is divided into 5 stages.
Because of its clinically relevant items, practical progression of starting positions, excellent interrater and intrarater reliability,11 and high predictive validity for outcome,8,11,12 the FM motor scale is commonly administered by physical therapists in clinical and research settings. Unfortunately, the time-consuming nature of the scale may limit its use.11,13 However, such a rank-ordered assessment scale may have the properties of a hierarchical scale; such properties would permit a more concise method of administration. All of the scale items would remain in the scale but would be tested only when relevant to the state of the patient.
A shortened version of the FM motor scale has been published by Hsieh et al.13 The 12-item scale (6 items for each section) has acceptable levels of reliability, validity, and responsiveness. The item selection for the shortened version was based on discussions of a group of experts along with the results of a Rasch analysis, from which only the scale items ranked highest and lowest for each subsection were selected. Unfortunately, this selection process excluded the scale item shoulder abduction and the hand subsection item finger extension. Because both of these scale items have been identified by Nijland et al8 as predictors of functional recovery after stroke, their potential inclusion in a scale is important. Likewise, Woodbury et al14 investigated the single (unidimensional) construct of the UE section of the FM motor scale by use of a Rasch analysis, which produced dramatic changes in the order of the scale items. In both studies, the stage divisions, which are an inherent part of the construct of the FM motor scale, were removed. Thus, the underlying conceptual framework of a stepwise hierarchy and the usefulness of the scale were compromised.
Rasch analysis, which is item based, requires a large and general sample. The strict rank order produced is based on probability statistics. An alternative, classic approach to rank ordering is Guttman scale analysis, which is based on deterministic ordering. Because it is based on the population of the sample rather than the items of the scale, Guttman scale analysis does not require such large samples or numbers of scale items. A minimum of just 3 items is required for Guttman scale analysis; this feature is useful for scales with fewer items and scales divided into subsections.15,16 The Action Research Arm Test, with 19 items and 4 subsections, is an example of a hierarchical Guttman scale. The Action Research Arm Test is widely used clinically and in research studies, partly because the time-saving administration protocol produces a valid total score.17,18
Guttman scale analysis of the FM motor scale has established stepwise hierarchical properties based on its unidimensional construct in patients with chronic stroke.19 When the scale is used as a Guttman scale, all of the items remain part of the scale; however, when a time-saving hierarchical protocol of administration is used, fewer scale items may actually need to be tested. Untested items can then be allocated scores to maintain a total score of 100 points. Furthermore, the established hierarchy supports the legitimate use of summed scores as a valid indicator of motor recovery. However, the results of a study of patients with chronic stroke cannot be automatically extrapolated to other stroke populations. Discovering that hierarchical properties are also applicable in acute stroke will support the use of the more concise method of administration of the FM motor scale in different phases of stroke. Therefore, the aim of this study was to repeat Guttman scale analysis in people within 72 hours after stroke to establish whether the same stage- and subsection-associated hierarchy of the FM motor scale exists in the acute phase.
Method
The present study is part of a prospective multicenter study conducted in the Netherlands (the Early Prediction of Functional Outcome After Stroke Study).20–23 Between May 2007 and May 2010, both the UE and the LE sections of the FM motor scale were administered within 72 hours after stroke onset as part of a multi-assessment battery.
Participants
A subcohort of 75 eligible people from an original database of 264 people who met the prospective multicenter study inclusion criteria was identified as an appropriate source of data for the present study. This subcohort included all people consecutively screened in the only hospital stroke unit at which the full version of the FM motor scale was administered as part of a multiassessment battery. All participants in the present study met the original study inclusion criteria (first ischemic stroke, monoparesis, or hemiparesis within 72 hours after stroke onset, 18 years or older, no disabling medical history, and no communication problems or cognition deficits that impeded measurement) and provided written informed consent. The eFigure shows the inclusion process. The 2 trained assessors were physical therapists, both of whom had more than 20 years of experience and were familiar with the measure. A 1-day training course was held to ensure correct interpretation of the administrative guidelines. At the completion of the training day, interrater reliability was informally assessed and found to be within the reported smallest detectable difference for the scale.24 All participants received treatment according to national rehabilitation guidelines,25 which are in agreement with current international rehabilitation guidelines.26,27
Assessment
The FM motor scale comprises 2 sections: the UE and the LE. The UE section consists of 33 scale items in 4 subsections: the shoulder-elbow-forearm (containing 18 items, divided into 5 stages), the wrist (5 items), the hand (7 items), and coordination (3 items). The LE section has just 17 scale items spread over 2 subsections: the hip-knee-ankle (containing 14 items, divided into 5 stages) and coordination (3 items). Guidelines indicate a specific scoring criterion for each scale item. The total motor scores for the UE and the LE are 66 points and 34 points, respectively. Reflexes are included in the total scores but are excluded from the analysis because they are considered to be inconsistent with a unidimensional motor construct.14,19
Data Analysis
Initially, the raw data were inspected to establish a spread of scores across all items in the stages and subsections as evidence that the participants had various motor abilities (Fig. 1). Evidence of a rank order from easy to more difficult items and a stepwise progression of motor recovery was then sought. First, progress that was not as expected from the rank order (indicated by failing scores on 3 items followed by a passing score on the next item) was identified as “nonscale type,” whereas motor recovery that progressed in the same rank order as that of the original scale was identified as “scale type.” Further inspection considered the stepwise progression of the stages in the shoulder-elbow-forearm subsection of the UE and the hip-knee-ankle subsection of the LE. For this process, the order of the stages was inspected to confirm that if a maximum total score was recorded for a stage, a maximum total score also was recorded for the preceding (easier) stage. Further inspection also confirmed that if no score was recorded for a stage, no further scores were recorded for subsequent (more difficult) stages. The rank order and stepwise progression were then similarly investigated for all of the remaining subsections. Furthermore, data for any participant who failed to score on any scale item or had passing scores on all scale items (floor or ceiling effect) was identified as “extreme.”
Spread of raw scores across all stages and subsections of the upper extremity and lower extremity sections of the Fugl-Meyer assessment scale.
The hierarchy of item difficulty was then investigated with the same process as that used in a previous study of chronic stroke.4 Guttman scale analysis15 was performed to confirm the hierarchical properties of the scale. The analysis was initially completed manually and later confirmed with stepwise permutations by use of Microsoft Excel 1997 (Microsoft Corp, Redmond, Washington). The raw data had to be transformed from a 3-point scoring system to dichotomous pass or fail scores by converting all scores of 0 to fail scores and combining scores of 1 or 2 points into pass scores. The alternative data transformation option (with cutoff points of a score of 0 or 1 being converted to a fail score and only a score of 2 being converted to a pass score) was also used. The first calculation for Guttman scale analysis was the coefficient of reproducibility (CR) because this value provides an estimation of the accuracy with which a total score for the scale predicts which items will have pass or fail scores. A CR of .9 or higher indicates a valid cumulative and unidimensional Guttman scale, such that scores can be legitimately summed.28 Next, the coefficient of scalability (CS) was calculated; the CS indicates the proportion of responses that can be correctly predicted from the total summed score. A CS of .6 is required to confirm the validity of Guttman scale analysis.28
The analysis process had 3 phases. In phase 1, we considered the hierarchical order of scale items in each stage of the shoulder-elbow-forearm and hip-knee-ankle subsections and in the remaining subsections of the UE and LE sections of the FM motor scale. In phase 2 of the analysis, we repeated the scaling procedure to investigate the hierarchical order of scale items between stages and subsections. In phase 3 of the analysis, we considered the hierarchy of scale item difficulty across all scale items when the stage divisions were ignored. Because Guttman scale analysis requires a minimum of 3 items, the analysis was not possible within stages 3 and 4 of the LE section; therefore, these were combined to create a joint stage, as in a previous study of chronic stroke.19 Finally, any change in the rank ordering of the scale items was examined to verify the allocation of appropriate scores for untested items, as recommended for the hierarchical administration of the scale.19
Role of the Funding Source
Data collection was funded by Wetenschappelijk College Fysiotherapie (WCF no. 33368) of the Royal Dutch Society for Physical Therapy (KNGF), Amersfoort, the Netherlands.
Results
Of the 75 eligible participants, 39 were men and 36 were women; the mean age was 67 years (SD=13, range=33–91). The hemiplegic side was on the left in 50 participants and on the right in 25 participants. According to the Bamford classification, 29 participants were classified as having a partial anterior circulation infarct, 26 were classified as having a total anterior circulation infarct, and 20 were classified as having a lacunar infarct.28 Most participants were assessed on day 1 (44%) or day 2 (35%) after their stroke.
Raw scores on the FM motor scale ranged from 0 to 66 (median=19, interquartile range=30) for the UE section and from 0 to 34 (median=19, interquartile range=14) for the LE section. No activity was seen in the wrist and hand in 43% of the participants, suggesting a moderately impaired participant group.29 Data for 3 participants were identified as extreme. One participant had failing scores on all items of both the UE and the LE sections (floor effect). Although a ceiling effect (ie, passing scores on all items of both the UE and the LE sections) was not seen for any participant, 1 participant scored maximum points for the UE section and another participant scored maximum points for the LE section. For the LE section, 7 participants had failing scores on 3 items followed by a passing score on the next item (nonscale type). Similarly, for the UE section, the findings in the shoulder-elbow-wrist subsection were nonscale type for 4 participants, and 12 participants had scores for items in the wrist or hand subsection or both but did not have scores for stage 4 in the shoulder-elbow-forearm subsection. Figure 1 shows the spread of raw scores. All extreme and nonscale-type data remained in the analyses as the error factors in the statistical formulas.
For all Guttman scale analyses, the results exceeded the recommended levels of acceptance of .9 for the CR and .6 for the CS.15,28 Table 1 shows the results for the 3 phases of analysis. The CR ranged from .95 to 1.00, indicating a valid cumulative scale and a unidimensional underlying construct. The range of .75 to 1.00 for the CS confirmed the validity of the hierarchical analysis. Because the analyses of both methods of data transformation produced equivalent values, only the results for the transformation with cutoff points of a score of 1 or 2 being converted to a pass score and a score of 0 being converted to a fail score are reported for ease of comparison with the results of a previous chronic stroke study.19
Guttman Scale Analysis of Upper and Lower Extremity Sections of the Motor Function Section of the Fugl-Meyer Assessment Scale
Inspection of the new rank ordering revealed some changes. However, these changes were only reordering of scale items to form a hierarchy within a stage. No reordering of items led to scale items being mixed between stages and subsections. For example, all of the scale items in stage 2 of the shoulder-elbow-forearm subsection appeared earlier (ie, easier) in the hierarchy than the scale items in stage 3. Likewise, inspection confirmed that if all of the items in stage 3 were tested and a score of 2 points was recorded, a score of 2 points also was recorded for all of the items in stage 2. These data support the allocation of 2 points for untested items. The findings of the present study indicate that the stepwise progression of the hierarchical order of the stages and subsections is correct. Tables 2 and 3 and the eTable show changes in the rank order in each subsection of the UE and LE sections of the FM motor scale as well as comparisons of these results with the original rank order3 and the hierarchical order in a previous chronic stroke study.19
Rank Order of Items in the Shoulder-Elbow-Forearm Subsection of the Upper Extremity Section of the Fugl-Meyer Assessment Scale
Rank Order of Items in the Wrist and Hand Subsections of the Upper Extremity Section of the Fugl-Meyer Assessment Scalea
Discussion
The results of the present study confirm the stage- and subsection-associated, cumulative, and unidimensional hierarchy of the UE and LE sections of the FM motor scale in people within 72 hours after stroke. Evidence-based practice requires accurate and valid tools to assess a patient's status and prognosis and to evaluate treatment. Therefore, because the more concise method of administration and the use of summed scores with such a robust assessment and outcome measure as the FM motor scale were found to be legitimate, their application should be encouraged.
The results of Guttman scale analysis in the present study were comparable to those in the previous chronic stroke study,19 as expected, because the spread of raw scores for people with acute stroke generally followed a pattern similar to that for people with chronic stroke. For the phase 1 analysis (within a stage or subsection), the range of CR values was slightly higher in the acute stroke study (present study). However, although all the results exceeded the recommended levels of acceptance for the CS, the lower range was just above the cutoff level of .6 in the chronic stroke study (at .61); the corresponding value in the acute stroke study was .83. This difference may be related to the diverse recovery patterns seen in stroke, such that different orders of item difficulty may occur within a stage or subsection. The phase 2 analysis (between stages and subsections) produced similar ranges in both the acute stroke study and the chronic stroke study. This finding was encouraging because it supported the stepwise nature of the hierarchy of the FM motor scale. Furthermore, the phase 2 analysis also supported the notion that it is appropriate to allocate scores for untested scale items.
The phase 3 analysis (across all scale items; stage divisions ignored) produced different results in the acute stroke study (present study) and the chronic stroke study.19 In the acute stroke study, a hierarchy was demonstrated by an acceptable level for the CR (.95), even when the stage divisions were eliminated. In the chronic stroke study, the CRs of .87 (UE) and .88 (LE) were below acceptable levels. However, the CSs in both studies were at acceptable levels (.75 [UE] and .65 [LE]). This difference may be related to the impact of time and the severity of the stroke on the course of recovery of motor control.7,30 Spontaneous recovery through the restitution of neural repair in the acute phase of stroke is replaced by substitution and adaptation through behavioral compensatory strategies in the later phases of stroke.5,30,31
Caution is still needed with regard to the hierarchy from the shoulder-elbow-forearm subsection to the wrist and hand subsections because the fact that fewer patients achieve wrist and hand recovery means that independent motor recovery of the wrist and hand remains under-investigated. Reordering of scale items occurred only within stages of the shoulder-elbow-forearm and hip-knee-ankle subsections and within the other subsections. With no changes in item order between stages or subsections, the stepwise hierarchy of the stages and subsections remained the same. These data indicate the correct grouping of scale items within the stages and subsections and the correct order of difficulty of the stages and subsections, forming a stepwise progression of motor ability. Therefore, no changes are required in the administration procedure, which remains a practical progression of tasks. This order also minimizes the number of changes in starting positions, saving time and energy.
The advantage of an assessment scale that can be administered hierarchically is that the measurement tool remains complete but does not need to be performed in its entirety. Furthermore, the starting level is set at a level appropriate to the patient, not necessarily at the beginning of a scale, if the scale items are obviously too easy. Therefore, potentially fewer scale items need to be administered; after 2 or 3 failing scores, the testing can be stopped. This process saves time and effort for both the patient and the clinician because only relevant scale items are administered. Through the use of the hierarchical properties of the scale, scores can be allocated to any untested scale items such that all scale item details are maintained. In this manner, summing of scores is legitimate, and the total score of 100 points for the FM motor scale is maintained, so that the scale can be used to indicate a patient's level of motor ability and prognosis. Furthermore, because scores from this qualitative motor impairment scale directly indicate a patient's level of recovery, this information can be used by a therapist to determine a treatment plan. Through the identification of the specific paresis and movement impairments, treatment strategies can be established to capitalize on motor ability and provide practice of movements and functional tasks at the appropriate level of recovery.1
Guidelines for the more concise hierarchical method of administration of the UE and LE sections of the FM motor scale have been published.19 To further assist in training on the more concise hierarchical method of administration, we developed decision flow diagrams for the UE and LE sections (Figs. 2 and 3). A recommended starting level is suggested for each extremity. The score on the initial scale item tested indicates the stage and scale item from which testing should continue. Links guide the user to allocate scores for untested items or identify specific scale items that need to be tested for scores to be awarded. With an appropriate starting level for the testing procedure, as few as 2 minutes are required to complete the process. Scale items from the wrist and hand subsections are always included in the hierarchical method of administration because the hierarchy for these subsections is less well established.
Flow chart for the hierarchical administration of the upper extremity section of the Fugl-Meyer (FM) assessment scale.
Flow chart for the hierarchical administration of the lower extremity section of the Fugl-Meyer (FM) Assessment Scale.
With the unidimensional hierarchy of the UE and LE sections of the FM motor scale established for both acute and chronic stroke, the next step is to confirm the robust nature of the FM motor scale. The population in the present study was predominantly elderly people who had stroke of moderate severity, were admitted to the hospital, and were assessed within 72 hours of their first stroke. Further work with a more diverse population of people with stroke is needed to confirm our findings and to add more information about the hierarchy into and within the wrist and hand subsections.
Footnotes
Ms Crow and Ms Harmeling-van der Wel provided concept/idea/research design. Ms Crow, Professor Kwakkel, Dr Bussmann, and Ms Harmeling-van der Wel provided writing. Mr Goos and Ms Harmeling-van der Wel provided data collection. Ms Crow provided data analysis. Ms Harmeling-van der Wel provided project management, institutional liaisons, and clerical support. Professor Kwakkel, Dr Bussmann, and Ms Harmeling-van der Wel provided consultation (including review of manuscript before submission).
The authors acknowledge the EPOS Investigators: Executive committee: Professor Gert Kwakkel, PhD, principal investigator, Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, the Netherlands, and Department of Neurorehabilitation, Reade Center for Rehabilitation and Rheumatology, Amsterdam, the Netherlands; and Barbara C. Harmeling-van der Wel, co-principal investigator, Department of Rehabilitation Medicine and Physical Therapy, Erasmus MC University Medical Center Rotterdam, Rotterdam, the Netherlands. Steering committee and data management: Janne M. Veerbeek, MSc, and Rinske H.M. Nijland, MSc, Department of Rehabilitation Medicine, MOVE Research Institute Amsterdam, VU University Medical Center, Amsterdam, the Netherlands. Monitoring board: Marijke A. van der Beek, UMC Utrecht; Utrecht, the Netherlands; Wam A.M. Cornelissen, AMC Amsterdam, Amsterdam, the Netherlands; Jos A.G. Goos, MNR, Franciscus Hospital, Roosendaal, the Netherlands; Claudia Steeg, UMC Sint Radboud, Nijmegen, the Netherlands; Richard Tichelaar, Amphia Hospital, Breda, the Netherlands; and Joke M. Timmermans, Leiden University Medical Center, Leiden, the Netherlands.
Ethics approval for the EPOS Study was granted by the Medical Ethics Testing Committee, VU University Medical Center, Amsterdam, the Netherlands.
Part of this study was presented as a poster at the World Confederation for Physical Therapy Congress; June 20–22, 2011; Amsterdam, the Netherlands.
Data collection was funded by Wetenschappelijk College Fysiotherapie (WCF no. 33368) of the Royal Dutch Society for Physical Therapy (KNGF), Amersfoort, the Netherlands.
- Received May 7, 2013.
- Accepted March 13, 2014.
- © 2014 American Physical Therapy Association