Metric properties of the ASIA motor score: Subscales improve correlation with functional activities¹ ☆

Article Outline

• Abstract
• Methods
• Results
• Discussion
• Conclusions
• References
• Copyright

Abstract 

Marino RJ, Graves DE. Metric properties of the ASIA motor score: subscales improve correlation with functional activities. Arch Phys Med Rehabil 2004;85:1804–10.

Objective

To apply item response theory (IRT) methods to neurologic and functional scales to determine the value of using American Spinal Injury Association (ASIA) motor subscores and ability estimates, rather than total ASIA motor scores, to predict motor FIM instrument scores.

Design

Secondary analysis of prospectively collected data.

Setting

Model Spinal Cord Injury Systems centers.

Participants

People with traumatic spinal cord injury (SCI) (N=4338) discharged from inpatient rehabilitation between January 1, 1994, and March 31, 2003.

Interventions

Not applicable.

Main outcome measures

Total discharge motor FIM scores, FIM subscale scores, and IRT-derived ability estimates of motor FIM scores.

Results

Use of separate ASIA upper-extremity and lower-extremity motor scores improved prediction of motor FIM scores over that of total ASIA motor score (R² for motor FIM score, .71 vs .59). Use of IRT-based ability estimates derived by applying a 2-parameter graded response model to the raw scores, however, did not improve prediction of motor FIM scores above that of the ASIA motor subscale scores.

Conclusions

Consistent with the metric properties of the ASIA motor score, and with recent models of disablement, impairment in SCI is more accurately characterized by using separate ASIA upper- and lower-extremity motor scores than by using a single motor score. Use of subscores for impairment should improve prediction of functional abilities and enhance more complex models of disability.

Keywords:  Activities of daily living, Factor analysis, statistical, Outcome assessment (health care), Rehabilitation, Spinal cord injuries

THE INTERNATIONAL STANDARDS for Neurological Classification of Spinal Cord Injury1 (the Standards) has become established as the primary neurologic description for traumatic spinal cord injury (SCI). The Standards were initially developed by the American Spinal Injury Association (ASIA) to classify neurologic deficit after SCI, and have since come to be used as an outcome measure. However, little research has been done to validate this use of the Standards. Examining patients according to the Standards yields several classification variables and scales. These include (1) single neurologic level; (2) sensory level, right and left; (3) motor level, right and left; (4) the ASIA Impairment Scale (AIS); (5) light touch score; (6) pinprick score; (7) motor score; and, for individuals with complete injuries, (8) extent of sensory and (9) motor zone of partial preservation, right and left. It is not clear how best to use these data. The ASIA motor scores have been most often used in describing recovery after SCI2, 3 or response to interventions.4 At times, motor recovery has been reported by subgroups based on AIS grade, such as complete (AIS grade A) versus incomplete (AIS grades B, C, and D),3 or by level of injury (tetraplegia or paraplegia).5, 6, 7, 8

Recent evaluations of the ASIA motor score suggest that it is not unidimensional. Factor analysis indicates that the ASIA motor score actually consists of 2 subscales, an upper-extremity motor score (UEMS) and a lower-extremity motor score (LEMS), each consisting of the respective 10 key muscles.9 Separating the ASIA motor score into these 2 scales extracts more information from the data, and has been shown to better predict changes in self-care and mobility during rehabilitation than use of a single motor score.9 The FIM instrument also has been shown to consist of 2 scales, a motor FIM and a cognitive FIM scale.10 Stineman et al11 have demonstrated that there are impairment-specific dimensions in the FIM and that, for traumatic SCI, the motor FIM scale should be divided into an upper cord FIM and a lower cord FIM.

These findings are consistent with recent developments in models of disablement. In Enabling America,12 the Institute of Medicine (IOM) describes a model of disablement and enablement in which disability is a result of an interaction between a person and the environment. In this model, pathology (eg, traumatic SCI) leads to an impairment (eg, muscle weakness), which results in a functional limitation (eg, inability to grasp). Whether this functional limitation leads to a disability depends on numerous factors both within and without the individual.13 The term functional limitation in this model refers to person-level ability to perform actions, rather than to the organ system focus of impairments. Functional limitations are the combined results of various impairments on the individual and should better reflect ability to perform activities than individual impairment measures. Lawrence and Jette14 used a model similar to the IOM model to determine the effect of functional limitations at 1 time point on subsequent development of disabilities. Their model divided functional limitations into “upper body” and “lower body” categories.

Recent models of disability, such as the IOM model and the revised World Health Organization (WHO) model, the International Classification of Functioning, Disability and Health,15 stress the importance of the environment in modifying the effects of impairments on activities. As a result, efforts are under way to evaluate the environment and to develop standardized measures of environment. People may function well in 1 environment, yet be more limited in another. A decrease in performance of self-care and mobility activities, as measured by the FIM, has been found in people with tetraplegia on discharge from initial inpatient rehabilitation.16 Level of function during inpatient rehabilitation is thought to more closely resemble capacity (what an individual can do under ideal circumstances) as opposed to performance (what an individual does in his/her usual environment, whether it be in the home or in the community).15, 17 The motor FIM in our study is used as a measure of “disability” (IOM model) or “activities” (WHO model). As such, it is summed scores that are relevant, not individual item scores. Although there are some floor effects for persons with high tetraplegia and some ceiling effects for persons with paraplegia, in general the total motor FIM score is a good basic measure of disability and it increases with lower levels of injury.18, 19

The purpose of our study was to use the IOM model of disablement to predict disability after traumatic SCI. Ours is an exploratory study of the relation between neurologic impairment, as measured by ASIA motor scores, and self-care and mobility activities, as measured by the motor FIM, in a standardized environment—namely, the inpatient rehabilitation unit. Results of models using raw scores are compared with models using item response theory (IRT) ability estimates. The hypothesis was that, by analyzing data using current standards of IRT for data elements, more precise predictions can be made. Specifically, we proposed that use of a separate UEMS and LEMS would better predict concurrent motor FIM scores than the total ASIA motor score, and use of ability estimates would improve predictions over use of raw scores. Minimizing variance in this restricted test situation—namely, the standard environment of inpatient rehabilitation facilities—facilitates the development of future models that add other factors, such as environment or social supports, to the mix.

Back to Article Outline

Methods 

This was a secondary analysis of data included in the National Spinal Cord Injury Database (NSCID), which contains data collected by the Model Spinal Cord Injury Systems (MSCIS) centers funded by the National Institute on Disability and Rehabilitation Research, US Department of Education. The research was approved by the institutional review boards of the respective MSCIS centers. Details of the origin and structure of the NSCID have been reported elsewhere.20 Discharge data from subjects with traumatic SCI, AIS grades A through D, discharged from inpatient rehabilitation between January 1, 1994, and March 31, 2003, were used in the analyses. To compare results using traditional raw scores with more modern approaches—that is, IRT-based ability estimates—data from subjects with incomplete discharge ASIA motor scores or motor FIM scores were eliminated.

Several neurologic variables were derived from the variables in the NSCID. Individual scores of key muscles were combined into a total ASIA motor score, the sum of the 20 key muscles, and 2 ASIA motor subscores: the UEMS and the LEMS, the sum of the 10 key muscles in the upper or lower extremities, respectively. A neurologic grouping variable was created to divide the sample into groups of persons with paraplegia and tetraplegia. The NSCID contains only the motor FIM items, not the cognitive items. A total motor FIM score was derived by adding the scores of the 13 motor FIM items. In addition, 2 motor FIM subscores were created, as recommended by Stineman et al11: an FIM upper cord score, consisting of the sum of the eating, grooming, bathing, dressing upper body, and walk/wheelchair item scores; and an FIM lower cord score, consisting of the sum of the remaining 8 items. Principal components factor analysis using orthogonal rotation was performed on the ASIA motor scores and motor FIM scores, to confirm the subscales.

IRT techniques were used to address deficiencies in classical use of the ASIA motor scores and motor FIM scores. For a more complete explanation of use of IRT techniques with ASIA motor scores, see chapter VI of the Reference Manual for the International Standards for Classification of Spinal Cord Injury.9 Briefly, the individual key muscles and the motor FIM items can be considered as items on a test. IRT specifies a mathematical model relating the probability of an item response to the magnitude of the underlying latent trait.21 In classical theory, there is no such specification, and all scores are linear combinations of item responses. Rasch models are 1-parameter models that estimate item difficulty but assume all items are equally discriminating. This means that a person’s function (ability) is determined by item score and the difficulty of the item. For motor FIM scores, for example, eating is easier than transfer from bed to chair. Therefore, a score of 5 for eating indicates less activity function than a score of 5 for transfers. The graded response model is a 2-parameter model that estimates both item difficulty and item discrimination. The ASIA motor scores and motor FIM scores were fit to a 2-parameter graded response model. The ability estimate in the graded response model is weighted by the difficulty of the items.22 This means that a person’s function (ability) is determined by the item score, the difficulty of the item, and how well the item discriminates among levels of the underlying trait.

The data were subsequently used to create maximum likelihood estimates (MLE) of ability, called ability estimates, of the trait underlying the scales. These MLE were established such that the distributions of the ability estimates have a mean of zero and a standard deviation of 1. IRT techniques allow evaluation of the amount of “information” each item contains regarding levels of the underlying trait. The more precise an item, the more information it contains. Items of different difficulty provide information about different levels of ability. Item information can be combined to determine test information, which indicates the range over which the test determines ability. The metric is called “Information”; values above 25 are desired. Graphing Information by ability for a trait provides a visual depiction of the range over which adequate estimates of ability are obtained. Such graphs were generated for ASIA motor ability and motor FIM ability.

Linear regression analyses were conducted to predict discharge motor FIM scores. Models were built in stepwise fashion, with forced inclusion of variables in each model. The first set of models predicted total motor FIM score from neurologic data. The first model included only total ASIA motor score to predict total motor FIM score. The second model added the neurologic grouping variable (tetraplegia vs paraplegia). The third model used the UEMS and LEMS to predict total motor FIM score. The fourth model added the neurologic grouping variable to model 3. The fifth model used the ASIA motor upper-extremity ability estimates (UEAE) and lower-extremity ability estimates (LEAE) to predict the total motor FIM ability estimate (FAE). A similar series of models was used to predict FIM upper cord scores and to predict FIM lower cord scores. Models were checked for multicolinearity. This occurs when predictor variables are highly correlated, and invalidates the estimate (β coefficient) of how much an individual item contributes to the prediction of the outcome variable.

Back to Article Outline

Results 

There were 4338 subjects with complete ASIA motor and motor FIM data at discharge from inpatient rehabilitation. Characteristics of the sample population at discharge are provided in table 1. The sample was predominantly men and white. Causes of injury were primarily vehicular crashes (43.9%), followed by falls (24.7%) and acts of violence (18.5%). More than half (53.5%) of the subjects had sustained tetraplegia, and nearly half (47.2%) had complete injuries at discharge from rehabilitation. The median age of the sample was 33 years, with an interquartile range (IQR) of 22 to 46 years. The median time from injury to rehabilitation admission was 15 days (IQR, 9–28d), and the median time in rehabilitation was 46 days (IQR, 29–73d).

Table 1. Demographic Characteristics of Study Sample

Figure 1 shows the distribution of scores for the ASIA motor and motor FIM scales and subscales. Median ASIA motor score at discharge was 50 (IQR, 31–70). The respective values for the UEMS and LEMS were 44 (IQR, 23–50) and 0 (IQR, 0–30). Forty-two percent of subjects had the maximum UEMS of 50, and 53% had the minimum LEMS of 0. A small percentage of subjects with paraplegia had a UEMS less than 50 as a result of non-SCI-related neurologic deficits, such as brachial plexus injuries. The motor FIM scores did not demonstrate these dramatic ceiling and floor effects. Figure 2 shows the distributions of the ability estimates for the scales involved. Note that the ability estimates tend to be more normally distributed than the raw scale scores. The exception is the UEAE and LEAE, because of the marked distributional abnormalities of the raw scores.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 1. 

  Boxplots of discharge ASIA motor scores and motor FIM scores. Legend: bars indicate median; boxes represent 25% to 75% range; lines indicate outliers. Abbreviations: FIMTOT, total motor FIM score; FIM_LO, FIM lower cord score; FIM_UP, FIM upper cord score; MS_TOT, total ASIA motor score.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 2. 

  Boxplots of discharge ASIA motor ability and motor FIM ability estimates. Ability is the mean likelihood estimate of ability. Legend: bars indicate median; boxes represent 25% to 75% range; lines indicate outliers; squares indicate range of extreme outliers. Abbreviations: FIM_LAE, FIM lower cord ability estimate; FIM_TAE, total motor FIM ability estimate; FIM_UAE, FIM upper cord ability estimate; MAE_TOT, total ASIA motor ability estimate.

Information functions for the ASIA motor ability estimates and the motor FAE are shown in Fig 3, Fig 4, respectively. Ability is plotted on the x axis, and information on the y axis. Ideally, the line should be above 25 on the y axis, although levels much greater than 25 indicate possible redundancy. Ranges of ability where the line is above 25 indicate ranges over which the scale measures ability well. Ranges of ability where the line is much below 25 indicate ranges where the scale does not measure the trait well. As seen in figure 3, separating the ASIA motor ability estimate into UEAE and LEAE subscales increased the range of ability estimates where information is greater than 25, particularly at the lower ability levels. In contrast, splitting the motor FAE into upper and lower cord subscales did not greatly improve the information function (fig 4).

• 
  • View Large Image
  • Download to PowerPoint

• Fig 3. 

  ASIA motor score information function. Legend: bars indicate range of ability estimates with desired levels of information: the striped bar represents total ASIA motor score values, the solid gray bar represents subscore values. Use of subscores increased the information at the lower levels of ability. Abbreviations: LEAE, lower-extremity ability estimate; MAE_TOT, total motor ability estimate; UEAE, upper-extremity ability estimate.

• 
  • View Large Image
  • Download to PowerPoint

• Fig 4. 

  Motor FIM information function. Legend: bars indicate range of ability estimates with desired levels of information: the striped bar represents total motor FIM ability estimate values, the solid gray bar represents subscore values. Use of subscores did not appreciably increase information.

Principal components (factor) analysis was consistent with the information function results. Factor analysis of the ASIA motor scores indicated that 2 factors accounted for 89.5% of the variance in the 20 scores. The items loaded on factors so that 1 factor consisted of the 10 upper-extremity muscle scores and the other factor consisted of the 10 lower-extremity muscle scores. Factor analysis of the motor FIM items identified only 1 factor that accounted for 91.5% of the variance in item scores.

Several models were used to predict total motor FIM score (table 2). The first model used only the total ASIA motor score. This explained 59% of the variance in motor FIM scores. Adding neurologic group (tetraplegia vs paraplegia) to the model (model 2) increased the model variance explained to 68%. However, there were multicolinearity problems in this model, making the β coefficients invalid. The third model used the UEMS and LEMS to predict discharge motor FIM. This model explained 71% of the variance in motor FIM scores. Adding neurologic group to model 3 did not increase the amount of variance predicted (model 4) and created multicolinearity problems. Model 5, which used ability estimates for both motor FIM scores and ASIA motor scores, did not increase the amount of variance explained above that of model 3.

Table 2. Regression Analyses to Predict Total Motor FIM Score

Abbreviations: CI, confidence interval; MS, ASIA motor score; NA, not applicable; Paraplegia, neurologic group; (tetraplegia or paraplegia); SE, standard error.

Results of models to predict the FIM upper cord scores are shown in table 3. There was greater variance explained by using separate UEMS and LEMS (72%, model 3) than by using the total ASIA motor score (44%, model 1) or by using the total ASIA motor score and neurologic category (60%, model 2). Again, adding neurologic group to model 3 did not increase the amount of variance predicted (model 4, data not shown). Models 2 and 4 had problems with multicolinearity because of the addition of the neurologic grouping variable. Use of ability estimates resulted in a slightly lower amount of variance predicted (65%, model 5) compared with raw UEMS and LEMS.

Table 3. Regression Analyses to Predict FIM Upper Cord Score

Results of models predicting FIM lower cord scores are shown in table 4. Model 3, using a separate UEMS and LEMS, explained 65% of the variance in FIM lower cord scores, which was equal to the variance explained by model 2 (using total ASIA motor score and neurologic group) and better than total ASIA motor score alone (60%, model 1). Model 4 did not improve the percentage of variance explained by model 3, and use of ability estimates for motor FIM scores and ASIA motor scores (model 5) resulted in a slightly lower percentage of variance explained. As with the other results, models 2 and 4, which included the neurologic grouping variable, had multicolinearity problems.

Table 4. Regression Analyses to Predict FIM Lower Cord Score

Back to Article Outline

Discussion 

The hypotheses for our study were partially confirmed. Using the UEMS and LEMS instead of a single ASIA motor score improved prediction of motor FIM scores. Adding neurologic category to the models improved prediction of motor FIM scores when total ASIA motor score was used, but not when the subscales were used. However, models with neurologic category were not valid models because of multicolinearity of the predictor variables. This problem did not occur when only the ASIA motor subscales were used.

Both factor analysis and IRT analyses support the separation of the ASIA motor score into 2 scales, an upper-extremity scale and a lower-extremity scale. These results confirm earlier studies evaluating predictive ability of the ASIA motor scores.9, 23 The ASIA motor score is not a true scale, and the same total score can be obtained with many combinations of scores of key muscles. For example, a score of 50 can be obtained by a person with complete paraplegia (UEMS=50, LEMS=0), by one with AIS grade C tetraplegia (UEMS=40, LEMS=10), or by one with an ASIA grade D tetraplegia central cord injury (UEMS=10, LEMS=40). These 3 people will likely have very different functional abilities. Separating the ASIA motor score into upper- and lower-extremity scores markedly reduces this problem. Therefore, use of total ASIA motor scores in analyses is discouraged.

The results of regression analyses suggest that ASIA upper-extremity motor function contributes more to the prediction of motor FIM scores than does ASIA lower-extremity motor function. The standardized regression coefficient (β) for UEMS and UEAE is larger than that for LEMS and LEAE in models 3 and 5 in all analyses Table 2, Table 3, Table 4. Although the UEMS and UEAE contribute about 3 times as much as the LEMS and LEAE in predicting total motor FIM, they contribute more than 10 times as much for FIM upper cord function, but only 1.5 to 2 times as much for FIM lower cord function. In contrast, Graves et al23 found that the UEMS contributes 3 times as much as the LEMS in the prediction of self-care gain during rehabilitation in traumatic SCI, whereas the LEMS contributed twice as much as the UEMS in predicting mobility gain. Graves used admission ASIA motor scores to predict change in motor FIM scores. Admission ASIA motor scores may be predictive of ASIA motor score gain, as suggested by data on motor score change based on initial muscle grade.5, 6, 7, 8 This may have given admission LEMS more predictive power for mobility gain during rehabilitation than ASIA motor scores at discharge. In addition, Graves did not use the Stineman-designated FIM upper cord and lower cord subscales, but used the existing self-care and mobility subscales, which likely contributed to the different results in these 2 studies. Further comparison of analyses, using ASIA motor scores to predict current function, future function, or change in function, will help make sense of these discrepant findings.

Our results are at odds with Stineman et al11 in that we did not find multiple dimensions in the motor FIM items. It is not clear why this was the case. The sources of data were different: Stineman used a large sample obtained from the Uniform Data System (UDS) for Medical Rehabilitation in 1992, whereas our data were limited to MSCIS centers from 1994 to 2003. Our sample may have differed from that of UDS data. Examination and classification of SCI is based on the Standards for all MSCIS, which may not be the case for all centers contributing to the UDS. Patients admitted to MSCIS centers differ from patients with SCI in population-based studies. A higher percentage of patients admitted to the MSCIS have complete injuries, more are injured by acts of violence, and fewer by motor vehicle crash than would be expected by population-based studies.24 Patients admitted to the MSCIS also differ from patients with traumatic SCI admitted to UDS-participating institutions. Using 1999 UDS data, subjects with traumatic SCI in the UDS database are older (mean age, 44y vs 36y), are admitted later after injury (mean, 24d vs median, 15d), and have shorter lengths of stay (mean, 34d vs 56d) than MSCIS patients.25 The UDS report does not give the level and completeness of traumatic SCI, so it is not possible to determine whether there are differences in impairment between these 2 populations. Further research is necessary to understand the differing results from factor analysis of the motor FIM in different populations.

There is a question of whether the motor FIM items measure the same trait and are appropriate for IRT analyses. It is possible that the order of item difficulty is not the same for different patterns of neurologic deficits. For example, eating may be easier than transfers for a person with complete tetraplegia, but harder than transfers for someone with a central cord syndrome. However, IRT methods, including Rasch analyses and the methods used in this study, have been applied to the motor FIM in the past, with acceptable fit statistics for these methods.19, 23 Evidence that items behave differently in different populations, called differential item functioning, has not been reported for the motor FIM, although this would be a good topic for future research. If differences were found, then there may be justification for dividing the motor FIM into subscales, if such subscales would correct the problem.

Unexpectedly, use of ability estimates instead of raw summed scores did not improve the prediction of motor FIM scores, as hypothesized. This may be because of the very nonnormal distribution of ASIA motor scores resulting from the discontinuous nature of the motor examination in the Standards. There are no muscles to test between the T1 myotome and the L2 myotome. It has been demonstrated that using sensory scores for these segments improves the prediction of functional gain.26 Unfortunately, sensory scores are not included in the NSCID. Other work using ability estimates of ASIA motor scores to predict changes in self-care and mobility function during rehabilitation has found that more variance is explained by ability estimates than by raw scores.9 Further study is needed to determine why, in this instance, no difference was found.

The UEMS and LEMS are impairment measures, not functional limitation measures. If upper- and lower-extremity functional limitation measures were available, these could improve prediction of functional abilities. Little work has been done to construct functional limitation measures for SCI. One measure of upper-extremity function, the Capabilities of Upper Extremity (CUE) instrument, has been evaluated in subjects with tetraplegia, but it requires further development.27 More recently, a lower-extremity measure to evaluate walking ability in SCI—the Walking Index for Spinal Cord Injury28 (WISCI)—has been developed. The CUE was found to be better than the UEMS in predicting motor FIM scores in people with chronic tetraplegia.27 Similar comparisons with the WISCI have not been done. Use of functional limitation measures may not have a great impact on prediction of function in isolated SCI, but should be more important in patients with comorbidities and secondary impairments, such as spasticity or contracture, which can limit functional abilities despite good strength on manual muscle testing.

Our study looked at motor and functional abilities at 1 time point—namely, discharge from rehabilitation—and in a standardized environment. Evaluation of metric properties of the ASIA motor score at other time points is indicated to confirm the 2-dimensional structure of this score. It would be interesting to look at the motor FIM at other time points as well, to see if the 1-dimensional structure is stable.

Back to Article Outline

Conclusions 

The ASIA motor score is actually 2 dimensional, consisting of a UEMS and an LEMS. Better prediction of functional abilities is obtained by using the scores of each dimension rather than the total score. When evaluating disability in SCI, these subscales should be used to determine the effect of impairment on function and to control for the effect of impairment when evaluating other influences.

Back to Article Outline

References 

1. American Spinal Injury Association . International standards for neurological classification of spinal cord injury, revised 2000, reprinted 2002. Chicago: ASIA; 2002;
   • View In Article
2. Ditunno JF, Cohen ME, Hauck WW, Jackson AB, Sipski ML. Recovery of upper-extremity strength in complete and incomplete tetraplegia (a multicenter study). Arch Phys Med Rehabil. 2000;81:389–393
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (52 KB)
   • CrossRef
3. Marino RJ, Ditunno JF, Donovan WH, Maynard F. Neurologic recovery after traumatic spinal cord injury (data from the Model Spinal Cord Injury Systems). Arch Phys Med Rehabil. 1999;80:1391–1396
   • View In Article
   • Abstract
   • Full-Text PDF (1048 KB)
   • CrossRef
4. Geisler FH, Dorsey FC, Coleman WP. Recovery of motor function after spinal-cord injury—a randomized, placebo-controlled trial with GM-1 ganglioside. [see comments] N Engl J Med. 1991;324:1829–1838
   • View In Article
   • MEDLINE
   • CrossRef
5. Waters RL, Yakura JS, Adkins RH, Sie IH. Recovery following complete paraplegia. Arch Phys Med Rehabil. 1992;73:784–789
   • View In Article
   • MEDLINE
6. Waters RL, Adkins RH, Yakura J, Sie I. Motor and sensory recovery following complete tetraplegia. Arch Phys Med Rehabil. 1993;74:242–247
   • View In Article
   • MEDLINE
7. Waters RL, Adkins RH, Yakura JS, Sie I. Motor and sensory recovery following incomplete paraplegia. Arch Phys Med Rehabil. 1994;75:67–72
   • View In Article
   • MEDLINE
8. Waters RL, Adkins RH, Yakura JS, Sie I. Motor and sensory recovery following incomplete tetraplegia. Arch Phys Med Rehabil. 1994;75:306–311
   • View In Article
   • MEDLINE
   • CrossRef
9. Graves DE, Marino RJ. Metric properties of the International Standards for Neurological Classification of Spinal Cord Injury, implications for research use. In: Neurological Standards Committee of the American Spinal Injury Association, editor. Reference manual for the International Standards for Neurological Classification of Spinal Cord Injury. Chicago: American Spinal Injury Association; 2003;p. 68–88
   • View In Article
10. Linacre JM, Heinemann AW, Wright BD, Granger CV, Hamilton BB. The structure and stability of the Functional Independence Measure. Arch Phys Med Rehabil. 1994;75:127–132
    • View In Article
    • MEDLINE
11. Stineman MG, Jette A, Fiedler R, Granger C. Impairment-specific dimensions within the Functional Independence Measure. Arch Phys Med Rehabil. 1997;78:636–643
    • View In Article
    • Abstract
    • Full-Text PDF (723 KB)
    • CrossRef
12. Institute of Medicine . Enabling America (assessing the role of rehabilitation science and engineering). Washington (DC): Natl Acad Pr; 1997;
    • View In Article
13. Verbrugge LM, Jette AM. The disablement process. Soc Sci Med. 1994;38:1–14
    • View In Article
    • MEDLINE
    • CrossRef
14. Lawrence RH, Jette AM. Disentangling the disablement process. J Gerontol B Psychol Sci Soc Sci. 1996;51:S173–S182
    • View In Article
    • MEDLINE
15. World Health Organization . International classification of functioning, disability and health (ICF). Geneva: WHO; 2001;
    • View In Article
16. Marino RJ, Cohen ME. Post discharge changes in the self-report functional independence measure in relationship to severity of injury. J Spinal Cord Med. 2000;23:29–30
    • View In Article
17. Marino RJ, Stineman MG. Functional assessment in spinal cord injury. Top Spinal Cord Inj Rehabil. 1996;1:32–45
    • View In Article
18. Hall KM, Cohen ME, Wright J, Call M, Werner P. Characteristics of the Functional Independence Measure in traumatic spinal cord injury. Arch Phys Med Rehabil. 1999;80:1471–1476
    • View In Article
    • Abstract
    • Full-Text PDF (722 KB)
    • CrossRef
19. Ditunno JF, Cohen ME, Formal C, Whiteneck GG. Functional outcomes. In:  Stover SL,  Whiteneck GG,  DeLisa JA editor. Spinal cord injury (clinical outcomes from the Model Systems). Gaithersburg: Aspen; 1995;p. 170–184
    • View In Article
20. Stover SL, DeVivo MJ, Go BK. History, implementation, and current status of the National Spinal Cord Injury Database. Arch Phys Med Rehabil. 1999;80:1365–1371
    • View In Article
    • Abstract
    • Full-Text PDF (930 KB)
    • CrossRef
21. van der Linden WJ, Hambleton RK. Handbook of modern item response theory. New York: Springer; 1997;
    • View In Article
22. Samejima F. Estimation of latent ability using a response pattern of graded scores. Psychometrika Monogr Suppl. 1969;34(4):100–114
    • View In Article
23. Graves DE, Frankiewicz RG, Carter RE. Gain in functional ability during medical rehabilitation as related to rehabilitation process indices and neurologic measures. Arch Phys Med Rehabil. 1999;80:1464–1470
    • View In Article
    • Abstract
    • Full-Text PDF (857 KB)
    • CrossRef
24. Go BK, DeVivo MJ, Richards JS. The epidemiology of spinal cord injury. In:  Stover SL,  DeLisa JA,  Whiteneck GG editor. Spinal cord injury (clinical outcomes from the Model Systems). Gaithersburg: Aspen; 1995;p. 21–55
    • View In Article
25. Deutsch A, Fiedler RC, Granger CV, Russell CF. The Uniform Data System for Medical Rehabilitation report of patients discharged from comprehensive medical rehabilitation programs in 1999. Am J Phys Med Rehabil. 2002;81:133–142
    • View In Article
    • MEDLINE
    • CrossRef
26. Graves DE, Frankiewicz RG. A sensory variable as predictor of functional gain. [abstract] Arch Phys Med Rehabil. 2000;81:1621
    • View In Article
27. Marino RJ, Shea JA, Stineman MG. The Capabilities of Upper Extremity instrument (reliability and validity of a measure of functional limitation in tetraplegia). Arch Phys Med Rehabil. 1998;79:1512–1521
    • View In Article
    • Abstract
    • Full-Text PDF (1343 KB)
    • CrossRef
28. Ditunno JF, Ditunno PL, Graziani V, et al.  Walking index for spinal cord injury (WISCI) (an international multicenter validity and reliability study). Spinal Cord. 2000;38:234–243
    • View In Article
    • MEDLINE

• 1 No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated.