Introduction

Measurements of functional outcomes are an integral part of any goal orientated, interdisciplinary rehabilitation programme and are important for quantifying the success of rehabilitation programme. The Functional Independence Measure (FIM) has gained widespread acceptance as a universal tool for functional assessment in rehabilitation. The FIM was designed to measure severity of disability and to determine ‘burden of care’ in core life activities across a range of impairments.1 It contains 18 items rated on a seven-point scale that reflect the amount of assistance a person requires in the areas of self-care, sphincter control, mobility, locomotion, communication and social cognition. The FIM has attained general acceptance largely because of its practical nature, reliability and ability to measure functional aspects of individuals’ overall independence.2 However, the validity of the FIM in the spinal cord injury (SCI) population has not been proven3 and it has some well-recognised limitations.4

The FIM was originally intended to measure burden of care, rather than level of functional independence, and was not specifically designed for people with SCI. It is therefore not surprising that there are some inherent problems with the FIM when used to demonstrate the success of rehabilitation programmes in the spinal cord injured population. In particular, the mobility and locomotor items of the FIM appear to lack sensitivity and the ability to distinguish important functional differences between individuals with different SCI impairments. This is reflected in the results of various studies that have failed to detect changes in function of patients over time and failed to detect differences in function of patients of various ages and neurological lesions. For instance, an early study by Menter et al5 found little variation in FIM scores between individuals with high level (T3–4) and low level (T11) paraplegia or between younger (<30 years) and older (>45 years) individuals. Dodds et al3 showed that although the FIM could discriminate between complete paraplegic and tetraplegic groups, the FIM was unable to demonstrate distinct differences along the continuum from high tetraplegia to low paraplegia, particularly when incomplete lesions were included. Similarly, Ota et al6 in their study of 100 patients with SCI found no differences in the FIM scores of persons with C8 tetraplegia, T1–5 paraplegia and T5–12 paraplegia, and Marino et al7 showed no improvement over a 3-month period following injury.

More recently, research by our own group,4 measuring functional outcome at discharge in 112 individuals with SCI, revealed that the FIM locomotor items were largely insensitive with marked ceiling and floor effects. In addition, the FIM locomotor items failed to delineate important functional differences between patients with different neurological lesions.

A number of alternative assessment tools have been proposed to assess functional independence more broadly. Particular emphasis has been placed on relating these assessment tools to individuals’ level of participation and activity (as specifically described in the International Classification of Functioning, Disability and Health – WHO, 2001).8 For instance, both the Spinal Cord Independence Measure (SCIM)9 and the Quadriplegic Index of Function7 have been proposed as alternatives to the FIM. Several other assessment tools, specifically designed to assess the mobility and/or locomotion function of wheelchair-dependent individuals, have also been designed (reviewed by Kilkens et al11). These include the Wheelchair Skills Test,10 Wheelchair Circuit11 and the Harvey Mobility Assessment Tool.12 While all the described assessment tools have merit, none have yet gained as widespread acceptance as the FIM nor been associated with such a well-organised, comprehensive education and accreditation programme. We therefore believed that time and effort may be better directed at improving the sensitivity of the FIM to detect important functional changes by adding a few simple FIM-like items, rather than advocating the use of different assessment tools.

Our aim therefore was to assess the responsiveness and discriminant validity of five additional mobility and locomotor (5-AML) items that could be used in conjunction with the FIM to enhance its sensitivity and better reflect relevant functional improvements with rehabilitation. We sought to determine the ability of the 5-AML to distinguish between individuals with varying neurological deficits (discriminant validity) and the ability of the 5-AML to detect changes in function over time (responsiveness). The 5-AML items were adapted from a six-item assessment tool of our own12 that was originally designed for wheelchair-dependent individuals with paraplegia. The original tool was devised by a group of Sydney physiotherapists experienced in the management of people with SCI. It was shown to have face validity with good interobserver reliability (weighted kappa 0.82–0.9812). The original tool has since been slightly modified to ensure that the items were appropriate for persons with tetraplegia and the scoring consistent with the FIM. While the initial assessment tool included a bed transfer item, this was removed as it measured the same domain as the existing three transfer items of the FIM. Hence, the five additional items that were trialed assessed the ability of wheelchair-dependent individuals with paraplegia and tetraplegia to move about on a bed, transfer from floor to wheelchair and push a manual wheelchair over level ground, ramps, and kerbs (see Appendix). It was hoped that by including both simple and complex tasks the scale would be able to better delineate important functional differences between groups with different levels of neurological impairment and would also help to eliminate the floor and ceiling effect of the FIM.

Methods

Subjects

Patients admitted to two acute spinal injury units in Sydney, Australia (The Royal North Shore Hospital and the Prince of Wales Hospital) between 1999 and 2002 were identified if they presented with a sudden onset of paraplegia or tetraplegia following trauma, surgery or a medical condition. The project was approved by the relevant hospital Human Research Ethics Committees and all participants signed an informed consent sheet. Inclusion criteria were persons aged between 18 and 65 years with a persisting neurological deficit following acute SCI. Patients likely to be transferred overseas within 6 months, patients with complete motor lesions at or above C4 and patients with ASIA Grade D impairments or motor function likely to enable the individual to walk, as well as those with multiple fractures were excluded from the study. One patient was excluded because of a psychiatric disorder. In all, 43 patients were included in the study.

Patients were tested within 72 h of mobilising in a wheelchair for the first time since injury, then 1 month, 2 months, 3 months and 6 months later (often after the person had been discharged from hospital). All but two subjects were assessed at one of the two acute hospitals or their affiliated rehabilitation centres. The remaining two subjects were unable to return to hospital for their 6 month assessment and consequently were self-assessed by telephone. In addition, data were missing for two subjects (both with paraplegia) at the 6-month follow-up. For these two subjects, data at 3 months were used to replace the data missing at 6 months. Data from another four subjects (all with paraplegia) were missing for two or more follow-up periods. These subjects were considered lost to follow-up and excluded from analyses. In this way, data from 39 of the original 43 subjects were included in the analysis.

Assessment procedure

At each assessment, subjects were scored using the ASIA sensory and motor examination,2 the locomotor and mobility subitems of the FIM13 and the 5-AML items. The ASIA examination is an accepted method of quantifying the neurological status of people with spinal cord injury and the FIM is a widely used measure of activity limitation for people with disabilities. Accredited staff of the two acute and rehabilitation units performed all FIM assessments. A research physiotherapist unaware of subjects’ FIM scores and experienced in SCI rehabilitation assessed subjects with the 5-AML items. The use of independent assessors to collect different aspects of the data reduced intraobserver bias.

Details of the 5-AML items

Tasks from an original mobility scale developed by Harvey et al12 were further refined and standardised. In particular, items were modified to ensure suitability for individuals with tetraplegia (see Appendix) and scoring was adjusted to a seven-point scale. The horizontal transfer item of the original scale was omitted because of its similarity with the FIM bed-transfer item.

The 5-AML items included two mobility and three locomotor items. One mobility item assessed patients’ abilities to get from a supine position in the middle of a gymnasium plinth (width approximately 108 cm) to a sitting position on the edge of the plinth in preparation for transfer. Patients were assessed on their ability to perform this motor task without an electric bed. The other mobility item assessed patients’ abilities to get from the floor back into their wheelchair. The three locomotor items assessed patients’ abilities to propel a manual wheelchair over flat ground, ramps and kerbs. Subjects were assessed in a manual rather than motorised wheelchair to validate the 5-AML locomotion items. The decision to assess subjects in manual, rather than motorised wheelchair, was based on the belief that the ability to mobilise in a manual wheelchair reflects an important aspect of independence. People with SCI often rely on a manual wheelchair when out in the community, especially if they depend for transport on vehicles unable to accommodate a motorised wheelchair. There is limited value in assessing individuals’ abilities to mobilise in a motorised wheelchair because mobility of people confined to electric wheelchairs is highly predictable and primarily reflects access to appropriate equipment.

Criteria for scoring the bed mobility and vertical transfer items were based on ability to complete the four key components of each task independently and within a reasonable amount of time (<3 min). Scoring criteria for the three additional locomotion items (ie pushing wheelchair on the flat, up and down a ramp, and negotiating kerbs) were based on a combination of factors, including standards for functional independence in the community (eg wheeling at ∼1.1 m/s to safely cross pedestrian crossing, ability to push up and down 1:14 gradient ramp, and negotiate 15 cm high kerb), practical distances and/or time limits, as well as a desire to discriminate effectively between different levels of impairment and physical capacity. With wheelchair pushing over level ground, for example, a decision was taken to set the highest score of ‘7’ (ie ‘complete independence’) for pushing wheelchair a distance of 200 m at a fast propulsion speed (ie >∼2.2 m/s) to avoid ceiling effects and acknowledge the fact that increased levels of physical capacity have been associated with reduced physical strain during activities of daily living (Janssen et al14) and a lower risk of medical complications, such as cardiovascular disease and development of pressure sores.

Statistical analysis

Subjects were classified by impairment into two groups, namely people with tetraplegia (C5–C8) and people with paraplegia (T1 and below). All subjects’ classifications remained constant over the course of the study. That is, their initial neurological classification did not change over the 6-month follow-up period. Median and interquartile ranges were derived for all item variables and represented in figure form as standard boxplots. Since data were ordinal, nonparametric statistics were employed. Construct validity was assessed by testing the ability of the 5-AML items to discriminate between neurological impairments (tetraplegia and paraplegia) using Mann–Whitney U-test at each time point. Item responsiveness over time was assessed by testing change over time (namely at 72 h, 1 month, 2 months, 3 months and 6 months after initial mobilisation) using the Friedman Test for repeated-measures. SPSS for Windows Version 1015 was used for all statistical analyses.

Exploratory factor analysis, as recommended by Streiner16, 17 and Gorsuch,18 was used to determine whether using the 5-AML items in combination with the FIM provided additional unique (or redundant) information. Factor analysis is a statistical procedure used to detect relationships between variables in a way that allows data to be reduced and intercorrelated variables to be grouped (or classified) into a limited number of common factors that explain the variability. Data at 1 month were used for this analysis because change in mobility and locomotor function was likely to be most pronounced at this time. Data from other time periods were then used to test the scale's responsiveness over time.

Prior to conducting the factor analysis, items that showed restriction in range were rejected. The ‘maximum likelihood’ approach was then used to estimate an initial ‘best’ representation of the relationship between variables, assuming that the observed variables and unobservable common factors have underlying multivariate normal distributions.19 The variance maximising (varimax) method20 was employed to reduce factorial complexity and assist easier interpretation, whereby the axes of initial factor scatterplot are rotated to distribute variance across factors more evenly and produce a pattern of loadings on each factor that is as diverse as possible. Eigenvalues of ‘1.0’ or more (accounting for equal or greater variance than is generated by one value) and visual inspection of the scree plot were used as criteria for retention of factors.21 The scree procedure (which plots eigenvalues for each item in descending order) provides a solution with the minimum number of factors accounting for the maximum amount of variance, whereby a clear break or change in the slope of plot curve is identified. Only items with a factor loading coefficient ⩾0.5 were included in each factor. Items were given equal weighting in the scoring procedure regardless of the size of the loading coefficient, as recommended by Gorsuch.18

Cronbach's alpha coefficient22 was calculated to establish the internal consistency of the items within each subscale and between each of the subscales. An alpha value of 0.6–0.7 was considered optimal, while a coefficient >0.9 was considered indicative of item redundancy.23 The construct validity of each subscale was examined by calculating the subscale scores and using nonparametric statistical tests24 to analyse differences between groups (tetraplegic and paraplegic) and within group differences over time.

Results

Table 1 provides information about the neurological status of subjects according to ASIA standards. All (n=39) but seven subjects were male. The median (interquartile range) age of subjects was 28 years (22–35 years). A total of 11 subjects were classified into the tetraplegic group and the remaining 28 into the paraplegic group.

Table 1 Characteristics of subjects (median and interquartile ranges) including age, sex and ASIA motor and sensory scores
Full size table

5-AML items

The results of the 5-AML items are presented in Figures 1 and 2. The bed mobility item (Figure 1a) demonstrates high responsiveness to change over time in the tetraplegic group, but quickly reaches a ceiling in the paraplegic group. The floor-to-chair transfer item (Figure 1b) discriminates effectively between the impairment groups with good responsiveness shown in the paraplegic group. All three locomotor items (Figure 2) differentiate between the tetraplegic and paraplegic groups. They show early responsiveness in the paraplegic group, although a later ceiling effect is reached for the flat push (Figure 2a) and ramp (Figure 2b) items for this group. The flat push and ramp items also show early responsiveness for the tetraplegic group. In contrast, the kerb item (Figure 2c) shows poor responsiveness for the tetraplegic group with a clear floor effect. The kerb item does, however, demonstrate good responsiveness to change over time for the group with paraplegia.

Figure 1
figure 1

Boxplots for mobility items of the 5-AML items. (a) Bed mobility item, (b) vertical transfer item

Full size image
Figure 2
figure 2

Boxplots for locomotor items of the 5-AML items. (a) Push on the flat item, (b) push on ramp item, (c) push on kerb item

Full size image

Figure 3 shows the distributions of the subscale scores for the 5-AML items. The subscale scores differentiated between the tetraplegic and paraplegic groups and demonstrated significant change over time (all Mann–Whitney U-test results, P<0.0001; Friedman tests, P<0.0001). These results demonstrate good construct validity. Cronbach's alpha coefficients for the subscale scores were 0.89 (5-AML items), indicating good internal consistency. No extremes or outliers were present for the paraplegic group for the 5-AML subscale scores and relatively few for the tetraplegic group.

Figure 3
figure 3

Boxplots for averaged score of the 5-AML items

Full size image

FIM mobility and locomotor items

The FIM mobility results are presented in Figure 4. The bed transfer item shows reasonable responsiveness over time for the paraplegic group but less so for the tetraplegic group. There is a clear ceiling effect of this item in the paraplegic group and a floor effect for the tetraplegic group. There appeared to be little difference between the three FIM mobility items for the paraplegic group.

Figure 4
figure 4

Boxplots for FIM mobility items. (a) Bed transfer item, (b) toilet transfer item, (c) bath transfer item

Full size image

The FIM locomotor items are presented in Figure 5. The items show severe restriction in range, both groups at all time points scoring either a 6 for the push/walk item (except for the tetraplegic group at baseline) or a 1 for the stair item. Neither item, therefore, discriminates between neurological impairment groups nor demonstrates changes in function over time.

Figure 5
figure 5

Boxplots for FIM locomotor items. (a) Push/walk item, (b) stair item

Full size image

Exploratory factor analysis

The three FIM mobility items and 5-AML items (eight items in total) were used for the exploratory factor analysis. The two FIM locomotor items were not suitable for use in this analysis because of the severe restriction in range. Although the study sample size was limited, an item to subject ratio of 1:5 was achieved, satisfying the minimum requirements for this type of analysis. The solution converged after eight iterations. Visual inspection of the scree plot and inspection of the initial eigenvalues indicated that a two factor solution was optimal, accounting for 80% of the variance.

The exploratory factor analysis identified a well-defined two-factor structure. The first factor to load contained the 5-AML items, with coefficients all above 0.63 (see Table 2). The second factor to load contained the three FIM mobility items, all loading with coefficients >0.86. These results suggest that there may be some item redundancy in the three FIM mobility items. Interestingly, the factor analysis indicated that it was appropriate to group the 5-AML items together.

Table 2 Rotated factor loading coefficient matrix
Full size table

Figure 6 shows the mean scores for the total scale using all eight items (ie, the 5-AML items and the three FIM mobility items) and demonstrates the utility of the whole scale for each group over different time points. The total scale shows adequate range across all time points for both groups, with occasional outlying values.

Figure 6
figure 6

Boxplots for averaged score of the three FIM mobility and 5-AML items

Full size image

Discussion

The FIM is a useful global measure in rehabilitation due to its well-understood psychometric properties, good reliability and broad applicability to a wide range of impairment groups.13 However, as a generic ‘gold standard’, evidence suggests it lacks some degree of specificity and sensitivity in certain functional domains for people with SCI. Previous research undertaken by our group4 and others3, 9, 25 has shown this to be most problematic for the locomotor, and to a lesser extent for the mobility items of the FIM. These limitations in the FIM need to be addressed, particularly in the current economic climate where funding is increasingly linked with outcomes.

Results of this study highlight the inherent problems of the two locomotor items of the FIM (ability to mobilise on flat ground and ability to ascend stairs) for wheelchair-dependent individuals (see Figure 5). Neither item is sensitive to changes in function and neither item distinguishes between individuals with different neurological lesions. This is not surprising as most wheelchair-dependent patients are unable to negotiate stairs but able to traverse 50 m (either in a manual or electric wheelchair). Consequently, people with SCI (and ASIA grade A–C impairments) typically attain a score of 1 for the stair item and 6 for the wheelchair locomotion item. Hence, scores on the two locomotor items of the FIM are highly predictable and do not measure real-life aspects of people's ability to mobilise: aspects that are responsive to rehabilitation. These observations were confirmed statistically.

Addition of the three new, more challenging locomotor items provided good discrimination between the two major SCI impairment groups studied and was highly responsive to changes in function over time. Measurement of ‘real world’ activities, such as pushing a manual wheelchair over prolonged distances and negotiating inclines and kerbs, reflects upper limb function, fitness and skill, and not merely access to appropriate equipment (namely, an electric wheelchair). The kerb item was particularly responsive to change over time in the paraplegic group. Interestingly, there was a ceiling effect on the flat push item in the paraplegic group despite the requirement to push faster than the normal speed of walking for a maximal score. The flat push and ramp items were quite sensitive to change over time in the tetraplegic group.

The three FIM mobility items demonstrated a clear ceiling effect for the paraplegic group and a floor effect for the tetraplegic group (Figure 4). In addition, improvements made by individuals in one item were mirrored by improvements in the other two, although there was some suggestion that bed transfers were somewhat easier to perform than either toilet or tub transfers. This reflects the similarity in the three skills (that is the ability to move horizontally with assistance of equipment albeit onto different surfaces). Redundancy in these FIM items was evident with statistical analysis. These findings are in accordance with previous research by ourselves4 and others,13 and raises the question of the additive value of measuring three rather than one FIM mobility item in the SCI population.

Our two additional mobility items were selected in an attempt to overcome the floor and the ceiling effect of the three FIM mobility items found in persons with tetraplegia and paraplegia, respectively. Hence, the inclusion of a simple (ie bed mobility) and more complex (ie floor-to-wheelchair transfer) motor task. The bed mobility item showed enhanced responsiveness over time for persons with tetraplegia, however, had very limited utility in the paraplegic group which rapidly reached a ceiling. In a similar way, the floor-to-wheelchair transfer item involving vertical lifting effectively discriminated between the impairment groups, demonstrating enhanced responsiveness and reduced ceiling effect for persons with paraplegia. In fact, preliminary analysis showed that the floor-to-wheelchair transfer item could also discriminate effectively between individuals with high (T1–6) and low (T7 and below) paraplegia.

The factor analysis revealed that a combination of the three FIM mobility items and the 5-AML items provided the best overall result, indicating that the 5-AML items were measuring a slightly different construct from the FIM. However, the subscales were highly correlated with one another (∼0.7) and one might question the value of administering both scales in this group of patients. Of course, a larger scale study is now required to validate these results and further refine the 5-AML items. It is generally agreed that studies of this nature should include 5–10 times as many subjects as there are items (in this case, up to 80 subjects26). However, fewer subjects may be acceptable if there is high loading of variables. This was the case in this study with several items loading at >0.8.27

The intention of the current paper was not to develop a new scale, but rather to explore the concept of enhancing certain less responsive domains of the FIM for individuals with SCI. This concept has been successfully adopted with the FAM in the traumatic brain injury population. While there are alternative scales that have recently been recommended for use in people with SCI, currently we believe that it is sensible to recommend a limited number of additional items that can be used in conjunction with the FIM to address some of its obvious limitations. The FIM is an internationally accepted and widely used tool. It adopts a simple scoring system that is readily understood. In certain domains, the FIM performs well for the SCI impairment group. For instance, there is an inverse relation between FIM self-care items and neurological level, as well as a consistent hierarchy of task difficulty.4, 28, 29 This study has shown that the 5-AML items, used in conjunction with the FIM, can improve the ability to discriminate between impairments. In addition, the 5-AML is easy and quick to administer. Its scoring system is adapted from that of the FIM, thereby minimising the amount of staff training required. However, like any assessment tool, staff will require some familiarisation with the items.

The 5-AML was designed to overcome the perceived problems of the FIM motor items when used to assess changes in motor function of wheelchair-dependent persons. It was not designed to assess changes in motor function of ambulant patients. The FIM may already have reasonable responsiveness for this subgroup. For this reason, patients with ASIA D impairments and the potential to walk were excluded.

Further studies with sufficient sample size are necessary to formally assess reliability of 5-AML items and further assess discriminant validity and responsiveness across adjacent neurological levels, as well as to perform a full factor analysis of 5-AML with all FIM items included. It may also be possible in future studies to correlate mobility and locomotor outcomes with intensity and type of rehabilitation. Finally, items of the 5-AML have not been weighted according to their difficulty, so that a maximum score of 7 is gained for both easier and more difficult or for frequently or less frequently performed items. It is acknowledged that further research is required to justify the meaningful summation of item scores.

Conclusion

The 5-AML items enabled better discrimination between people with tetraplegia and paraplegia. The inclusion of simple and more difficult tasks also enhanced scale responsiveness with detection of clinically important changes in function over a 6-month rehabilitation period. The results of this study indicate that the three FIM mobility items used in conjunction with the 5-AML items provided an improved measure of mobility and locomotor function in wheelchair-dependent people with SCI than the FIM alone. These combined results provide strong support for construct validity of the 5-AML items.