| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Movement Continuum Theory |
DD Allen, PT, PhD, is Adjunct Associate Professor, Department of Physical Therapy, Samuel Merritt College, Oakland, Calif, and Post-Doctoral Fellow, Health and Disability Research Institute, Boston University, Boston, Mass
Address all correspondence to Dr Allen at: allendianed{at}gmail.com
Submitted July 12, 2006;
Accepted March 1, 2007
 | Abstract |
|---|
Subjects: Thirty-five adults starting outpatient physical therapy intervention completed the MAM at the initial visit, at 2 weeks, and at 2 months or discharge. Thirty-four no-intervention volunteers completed the MAM twice.
Methods: The MAM responses were analyzed with item response theory methods; t tests were used to compare responses across test occasions.
Results: Paired t tests revealed significant changes in the intervention group at both 2 weeks and 2 months, with an effect size of 0.90 and a responsiveness index of 5.62 at discharge. At 2 weeks and at discharge, 57% and 80% of participants, respectively, showed gains greater than the minimal clinically important difference. Participants in the no-intervention group showed no significant change.
Discussion and Conclusion: The MAM responses revealed significant and clinically important changes following intervention. The MAM shows promise as a self-report measure of the effectiveness of physical therapy intervention.
| Introduction |
|---|
 TopAbstract Introduction MethodResultsDiscussion and ConclusionReference |
|---|
The MAM is a self-report measure created with a focus on movement, a concept common to broad diagnostic groups and specific to the work of physical therapists. The MAM has its foundation in the Movement Continuum Theory (MCT) of physical therapy2 and in a multidimensional extension of the MCT (see the article by Allen on 6 dimensions within the construct of movement in the Movement Continuum Theory in this Special Series). The MCT posits that people have a maximum achievable movement potential and a current and preferred movement capability. The multidimensional extension of the MCT specifies the idea of movement further by subdividing the construct of movement into 6 dimensions: flexibility, strength, accuracy, speed, adaptability, and endurance. The MAM puts the MCT and the 6 dimensions of movement into operation by asking people to record how they think they move now and how they would like to be able to move in the 6 dimensions. If people's perceptions of movement ability change with an episode of care, then their responses on the MAM should change.
Assessing how well an instrument detects change involves different criteria, depending on the purpose of the instrument.3 Instruments designed for longitudinal testing in effectiveness research should reflect significant and meaningful changes between measurements obtained before intervention and measurements obtained after intervention. Evaluating such instruments involves examination of sensitivity to change and responsiveness; the definitions of these terms overlap.3 Sensitivity to change means that the instrument can accurately detect change when it occurs, a concept predicated on the reliability of the instrument. That is, different responses over different occasions should reflect true differences in ability rather than measurement error. Measures of sensitivity to change include effect size, standardized response mean (SRM), and paired t values for single-group comparisons.3 Responsiveness incorporates the concept of sensitivity to change but also involves comparison of changes in a group with changes in a source external to the group, such as changes in a no-intervention control group. Measures of responsiveness include the responsiveness index4 and independent t values.3
Even with evidence of sensitivity to change and responsiveness, the changes recorded may mean very little to the physical therapist or to the patient or client.5 The meaningfulness or clinical value of reported differences following intervention depends on additional criteria, which are sometimes reported as the minimal clinically important difference (MCID)4 or the minimal important difference.6 Calculating the MCID helps to quantify the size of the difference necessary to have meaning for the clinician or patient. For the clinician, meaningful differences may signify a turning point in the course of the intervention; for the patient, meaningful differences may signify a new freedom to resume some previous activities.5 In a study in which the Berg Balance Scale was used to record balance changes with intervention in participants with multiple sclerosis, a 1-point difference out of a total of 56 points reflected insufficient change for the participants to find it meaningful; a 6-point change was designated the minimum needed to have clinical importance for the participants.7
Determination of the MCID relies on either normative or criterion-based methods.6,8 In a normative method, the MCID is based on the distribution of scores within the group tested. This method equates the MCID with the standard error of measurement (SEM), which is mathematically related to the standard deviation of the scores at baseline and the reliability of the scores obtained with the instrument. In a criterion-based method, the MCID is anchored on some external source. Possible external sources include therapist experience with a particular measure or consensus with other therapists, achievement of intervention goals or therapist-recognized improvement, or patient statement or response on a concurrent measure, such as a global rating of difference scale.8 Once calculated, the MCID may serve as a standard against which to compare changes in an individual or as a cutoff point for proportions of people in an intervention group who achieve the MCID and the proportions of people who do not.5
For this study, multiple criteria served to help interpret and place into context1 changes noted in MAM responses following intervention. Evaluation of sensitivity to change incorporated calculation of effect size, calculation of SRM, and single-group t test comparisons. Evaluation of responsiveness incorporated calculation of the responsiveness index and other comparisons with a no-intervention group and with criteria obtained from the attending physical therapists regarding the success of therapy. Attending physical therapists also completed the MAM for each participant in the intervention group after the initial examination to provide an additional external source against which to compare the participants’ self-reported movement ability. Interpreting the meaning of different responses following intervention involved calculation of the MCID and comparison of the proportions of participants in different groups who achieved the MCID. Calculation of the MCID relied on a normative method because no previous experience with the MAM could be used to establish a criterion anchor.
| Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussion and ConclusionReference |
|---|
|
Recruitment for the intervention group proceeded via verbal invitation when patients began a course of physical therapy intervention at participating outpatient clinical sites. Participants in the intervention group completed the MAM once at the beginning of physical therapy intervention (time 1), again at about 2 weeks into the episode of care (time 2), and again at 2 months or discharge, if the patient was discharged before 2 months had passed (time 3). Participants were included in the analysis if they completed at least 2 MAMs. In addition to the MAMs completed by the participants, each attending physical therapist completed the MAM once after the initial examination (time PT) for each participant in the intervention group. At the end of the episode of care, the attending physical therapist also determined the success of physical therapy intervention and recorded the answer. The training of attending physical therapists consisted of written and verbal instructions to complete 2 tasks: to complete the MAM after the first visit and to indicate at the end of the discharge visit whether physical therapy intervention for the participant was completely successful, partially successful, or not successful (therapists were instructed to choose one). All participants were informed that completing and returning the questionnaire constituted consent for their (anonymous) responses to be included in the study.
Forty-three participants began the study at 9 different outpatient clinics with 12 different attending physical therapists; 35 completed at least 2 MAM questionnaires. The average age of the 35 participants in the intervention group was 47 years, with a minimum and a maximum of 19 and 85 years, respectively. Twenty-five participants were female. The reasons for physical therapy intervention in this group included low back pain; neck or upper back pain; shoulder pain; knee, wrist, or foot surgery; ankle tendinitis; carpal tunnel syndrome; and thoracic outlet syndrome. Of the 12 attending physical therapists, 1 had a professional doctoral degree (at least 2 have since earned transitional doctor of physical therapy degrees), 1 had a certification in orthopedic manual therapy, 5 were owners of their clinics, and 2 were managers of corporation-owned clinical facilities. Years of physical therapy experience ranged from about 2 years to over 20 years.
Thirty-six volunteers from the community completed the MAM twice, at about a 2-week interval. Two were undergoing or about to start physical therapy intervention and so were removed from further analysis for the no-intervention group. The average age of the remaining 34 participants in the no-intervention group was 54 years, with a minimum and a maximum of 19 and 78 years, respectively. Seventeen participants were female.
After recruitment, participants in the intervention group received physical therapy intervention as directed by their therapists. Attending physical therapists managed the movement disorders of people in the intervention group without specification or constraint from this study, and intervention records were not collected. If people perceived an improvement in their movement ability with intervention, then they were expected to choose responses at a higher level at time 2 or time 3 than at time 1. Gains on the MAM were expected to be greater for people who had completely successful courses of physical therapy than for those who had partially successful or unsuccessful courses, as designated by their attending physical therapists. People in the no-intervention group were not expected to change their responses on the MAM.
Responses on the MAM were analyzed with item response theory (IRT) methods and ConQuest software.10,* Other sources provide an overview of IRT methods and the associated terminology11–13; therefore, only summary statements appear here. With IRT methods, the responses of all participants to the 24 items are converted into the probabilities of those responses and the natural log of the odds (logits) of each participant responding in a particular way to these items. The resulting logits designate the location of each item and participant on an interval scale. The scale for the MAM is the self-perception of movement ability in logit units.
Reliably estimating item locations on a logit scale requires a fairly large heterogeneous sample so that participants’ responses cover all actual response levels for all items. Such a requirement could restrict clinical studies to prohibitively large and unrealistically heterogeneous samples. However, IRT methods provide the possibility of anchoring item locations on previous calibrations. With item locations anchored, participant locations can be estimated and compared, even with small and relatively homogeneous samples. For this study, all analyses were performed with the item locations calibrated from a previous study.9 The previous study included 318 community-dwelling adults representing a heterogeneous mixture of movement ability; about 10% of the adults were starting outpatient physical therapy intervention. The logits for the levels of item responses (technically, item steps, or boundaries between 2 response levels) in the calibration study ranged from –7.7 to +9.8. The mean for the calibration population was set at 0 logit; the standard deviation was unconstrained. Standard errors ranged from 0.7 logit for a few of the lowest levels of responses to items to 0.1 logit for most of the highest levels of responses to items. The item separation reliability for the calibration study was .98.10
With item response level locations anchored on these previous calibrations, participant locations in this study were estimated and differences were calculated in 4 pair-wise comparisons of test occasions: from test to retest, time 1 to time 2, time 1 to time 3 (or discharge), and time 1 to time PT. Within each pair, the first test occasion was the baseline. For 6 participants in the intervention group who were discharged at 2 weeks, the completed time 2 MAM response sets also were counted with the time 3 MAM response sets in the analyses comparing baseline and discharge data.
Assessing the sizes of differences within each pair-wise comparison involved 2 procedures. In one procedure, the main effects for each pair were computed in ConQuest10 and assessed for significant differences. In the other procedure, each pair was examined with paired t tests. For computing the main effect, test occasion could be modeled as a main effect within the item response model for each pair-wise comparison. ConQuest produced estimated logit values and standard errors for each pair, and the differences within pairs were observed. The main effect of test occasion was deemed significant if the difference within the pairs of estimated logit values was greater than 1.96 times the standard error.10 For examining the results of paired t tests, differences in participant location estimates over time were calculated for each participant by subtracting the estimated baseline location from the designated comparison location. Average differences across participants were assessed with paired t tests. One-tailed analyses were performed for time 1-to-time 2 and time 1-to-time 3 comparisons because of the anticipated increase in perceived movement ability following intervention. Two-tailed analyses were performed for test-to-retest and time 1-to-time PT comparisons. The test-to-retest 
value was set at .05. Because the same baseline data were used for the other 3 pair-wise comparisons, the Bonferroni correction for 3 comparisons14 converted the value to .017.
Effect size14 was calculated as the average difference in the locations of participants from baseline to comparison occasion divided by the standard deviation of the distribution of baseline locations. The SRM3 was calculated with the same numerator divided by the standard deviation of the distribution of differences in locations from baseline to comparison occasion. The responsiveness index4 was calculated with the same numerator divided by the standard deviation of the distribution of differences in locations from baseline to comparison occasion for the no-intervention participants. Effect size thus relates change to the initial variability in the same group, SRM relates change to the variability of changes over test occasions in the same group, and the responsiveness index relates change to the variability of changes over test occasions specifically in the no-intervention group.
The MCID was calculated with the equation:
|
|
is a measure of internal consistency, a type of reliability.
Comparing the MAM responses of intervention group participants and attending physical therapists was a form of interrater reliability testing; therefore, an additional calculation, the intraclass correlation coefficient [ICC(2,k)], was carried out with SPSS software (version 11.0)
to examine their relationship. Intraclass correlation coefficient model 2 was selected so that results could be generalized to different therapists who might treat a participant during an episode of care. Intraclass correlation coefficient form k=24 was selected because the locations of participants were estimates based on responses to 24 items.
The intervention group was further subdivided into groups based on whether the attending physical therapist deemed the intervention completely successful, partially successful, or not successful. Average gains for each of the smaller groups were calculated and compared with 1-tailed independent t tests.
| Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussion and ConclusionReference |
|---|
Paired t tests revealed that gains in the locations for participants who underwent outpatient physical therapy averaged 0.98 logit at 2 weeks, equivalent to about 7 raw score points for the average participant (n=35, P<.00002). Gains in the locations for participants averaged 2.53 logits at 2 months or discharge, equivalent to about 17 raw score points for the average participant (n=25, P<.00005). In comparison, participants in the no-intervention group showed no significant change in locations, with an average difference of –0.10 logit after approximately 2 weeks (n=34, P=.22). In contrast to the results obtained with the main effects procedure, paired t tests revealed no significant difference between time 1 and time PT locations (n=34, P=.27).
With a change of 2.53 logits at 2 months or discharge, the effect size was 0.90, the SRM was 0.93, and the responsiveness index4 was 5.62 (Tab. 1). The effect size was much smaller (0.38) after only 2 weeks of intervention.
|
of .94 calculated for the MAM (see the article by Allen on the validity and reliability of the Movement Ability Measure in this Special Series). Twenty of 35 participants (57%) with time 1 and time 2 data showed a change on the MAM equal to or greater than the MCID at 2 weeks. Twenty of 25 participants (80%) with time 3 or discharge data showed a change on the MAM equal to or greater than the MCID at 2 months or discharge. For participant locations estimated from participant time 1 data and from physical therapist time PT data (n=34), the ICC(2,24) was .78. Locations resulting from physical therapist perceptions of participants’ movement ability averaged 0.44 logit lower than locations resulting from participants’ perceptions of their own movement ability. Physical therapists’ responses across items were much less variable than participants’ responses on the MAM (variances of 4.81 and 9.80, respectively).
For the 25 participants for whom a physical therapist assessed the success of the intervention, a 1-tailed t test revealed a significant difference (P<.05) between those assessed as having a partially successful intervention (n=12, average gain of 1.32 logits) and those assessed as having a completely successful intervention (n=13, average gain of 3.30 logits). No physical therapist designated a course of intervention as not successful for any of these participants.
| Discussion and Conclusion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussion and ConclusionReference |
|---|
Although the MAM was able to record change in the intervention group, potential biases limit the generalizability of the changes observed here. Participants in both groups were volunteers, possibly having a bias toward interest in movement ability. In addition, participants in the intervention group had multiple physical therapy intervention sessions, which would help focus their attention on movement, thereby enhancing this potential bias in a way not duplicated in the no-intervention group. Participants who completed the study may have had a positive bias toward physical therapy intervention; those who did not have a positive bias or who were initially skeptical about physical therapy intervention could have dropped out of the study or refused to start. Also, participants who completed the study may have experienced greater changes along the way than, for example, the 8 participants in the intervention group who dropped out before completing a second MAM. The physical therapists’ bias toward greater change is revealed by the fact that no physical therapist judged a course of intervention not successful for participants who completed the study. Physical therapists may have overestimated the success of the intervention, because for 3 of 13 participants judged to have partially successful therapy and for 2 of 12 participants judged to have completely successful therapy, there was a difference between baseline and discharge that was less than the MCID.
To examine further the possibility of participants’ bias toward change, data and gains at times 2 and 3 were analyzed. Of the 15 participants who showed a change of less than 0.61 logit (MCID) at 2 weeks, 3 participants were specified as being discharged at that point, 6 participants had no time 3 data, 2 participants continued but still showed a change of less than 0.61 logit at 2 months, and 4 participants continued and showed a change greater than the MCID at discharge. In comparison, 5 of the 20 participants who showed a change greater than the MCID at 2 weeks had no time 3 data. Therefore, dropping out of the study occurred among both participants who were making good progress and those who were not. Replication of this study is needed to determine whether the apparent difference in proportions of participants dropping out (6 of 15 showing less change and 5 of 20 showing change equal to or greater than the MCID) is significant or whether similar changes recorded with the MAM across occasions occur in other samples.
Note that any changes recorded in this study cannot be attributed definitely to the effect of physical therapy intervention, although the time course coincided with an episode of care. Nor can the lack of change in MAM responses for the no-intervention group be attributed definitely to a lack of change in movement ability for this group, although no intervention took place. A randomized controlled trial with the MAM as a dependent variable could help to establish the usefulness of the MAM in providing evidence of the effectiveness of a particular intervention compared with the lack of that intervention.
Although the ConQuest modeling method indicated that the main effect of test occasion was significant between time 1 and time PT, the paired t test for this comparison revealed no significant difference. The ICC for this comparison was relatively strong, at .78. The mixed results indicated that there probably were some differences in how physical therapists and participants judged movement ability but that their judgments were fairly strongly correlated. The time PT data would provide stronger evidence supporting the accuracy of movement ability reports from participants if not for large differences in variances for the 2 groups. The physical therapists tended to mark one level across most items on some of their completed MAM questionnaires, possibly reflecting time constraints, with a resultant restriction in variability unmatched by participants’ responses. Participants tended to perceive their own movement ability at a higher level than their therapists did at the beginning of the course of physical therapy intervention. Future studies could use performance-based data or therapists’ written examinations for comparison with patients’ MAM responses to determine whether the patients’ perceptions of change in movement ability have corollary evidence to support them.
Few clinicians have the expertise locally available to run IRT analyses of data gathered from their patients. The classical scoring method of adding up points across items on a measure like the MAM can provide some indication of people's perceptions of movement ability. Performing statistical analyses and comparisons such as those used in this study, however, assumes that the underlying scale, like the logit scale, has the same intervals between units. The units in a raw score do not meet this criterion in measures such as the MAM, in which item response levels are ordinal. To facilitate clinical utilization, Table 2 shows the conversion of some raw scores into logits on the basis of a previously performed calibration study. The raw scores for the intervention group at time 3 in this study averaged about 96; the raw scores for the no-intervention group at both test and retest averaged about 109.
|
In this study, the MAM reflected changes that were both statistically significant and clinically meaningful. The MAM shows promise as a way to document gains in perceived movement ability across a population receiving outpatient physical therapy intervention. Generalization to diagnostic groups other than the ones represented here or settings other than outpatient clinics will require further research.
| Acknowledgments |
|---|
The Committee for the Protection of Human Subjects at the University of California, Berkeley, designated this study exempt from further review.
Data and constructs of this manuscript were presented in a platform presentation at the International Objective Measurement Workshop; April 6, 2006; Berkeley, Calif.
This article arose from the author's doctoral dissertation at the University of California, Berkeley, December 2005.
| Footnotes |
|---|
SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.
| Reference |
|---|
|
TopAbstractIntroductionMethodResultsDiscussion and ConclusionReference |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
