Reliability of measurements of static weight distribution of manual wheelchairs.

The complexity of the clinical decisions related to manual wheelchair recommendation has increased tremendously during the past decade. More manufacturers are producing an increasing number of models of manual wheelchairs with a greater variety of available components. Additionally, health care practitioners must decide how wheelchair components should be configured for maximum user benefit. A critical decision is the position of the rear axle. Brubaker[1] reviewed mechanical and biomechanical factors affecting wheelchair propulsion performance and stated that the position of the rear axle with respect to the combined center of mass of the wheelchair and user determines the stability and maneuverability of a wheelchair. He"suggested that to increase propulsion efficiency, the percentage of weight distributed to the rear wheels should be increased for most wheelchair users. There is a concomitant increase in "tippiness" of the wheelchair for the user when this is done. Changes in the rear axle position are usually made based on the clinician's experience, the wheelchair user's experience, and the results of time-consuming trials. We believe there is a tendency to opt for more stable rather than more efficient configurations in the absence of measures of stability, propulsion efficiency, or weight distribution.

Although there has been no report of a clinically practical means of collecting weight distribution data, two approaches to the measurement of related characteristics of occupied wheelchairs have been reported. Loane and Kirby[2] developed a tilting platform to determine the angle of inclination at which an occupied wheelchair would become unstable and tip posteriorly. Kirby et al[3] also reported a method using an adapted goniometer to measure the angle at which the wheelchair and user are balanced on the rear wheels. The measure served as an indicator of the angle of inclination at which a wheelchair and user would tip posteriorly. Lemaire et al[4] reported specifications for a balance platform that can be used to determine the horizontal distance from the rear axle to the center of pressure of a wheelchair and user. This distance can be used, along with other data, to calculate an estimate of the rolling resistance for a wheelchair and user. These techniques may be in limited use clinically because the technology is not readily available or because there may be little perceived usefulness of the information they provide.

One commercially available balance system can be used to directly measure the weight distribution among the wheels of an occupied wheelchair. There have been no reports of the accuracy or reliability of measurements obtained with the system. The purpose of this study was to assess the reliability of measurements obtained with this system for the static distribution of vertical force (weight) among the four wheels of an occupied wheelchair.

Method

Instrumentation

The Chattecx Balance System[TM](*) was designed to measure responses of upright humans to perturbation of balance. The system (Fig. 1) consists of two footplates, signal conditioning circuitry, a personal computer with monitor and printer, and a platform that rotates about a horizontal axis or moves linearly. Each footplate has two force transducers connected by a sliding mechanism for length adjustment. The force transducers are 10.1 cm wide and can be separated to allow for independent positioning. The force transducer designed for placement under the forefoot is 15.3 cm long, and the transducer for the heel is 9.2 cm long. Each of the four transducers has a reference line that is used to align the transducer to the foot during balance testing. The following description of the instrumentation is that supplied by the manufacturer. Each force transducer has a set of full-bridge strain gauges. The analog transducer signals are amplified and digitally sampled at 100 Hz by the computer for analysis. Each force transducer is calibrated to zero when the system is turned on and can measure up to a maximum of 54.43 kg (120 lb). The system is designed to calculate a maximum total load of 170.1 kg (375 lb). The instantaneous center of pressure, relative to the four force transducer positions, is calculated and displayed on the monitor. For the purposes of our experiments, all testing was done with the force transducers placed on a level tile floor and with a 10-second sampling period.

Procedure

Our study consisted of three experiments. In experiment 1, calibration weights of known mass were repeatedly placed on the force transducers to determine the amount of variation that could be expected in measurements obtained with the system without a change in the load on the force transducers. In experiment 2, unoccupied wheelchairs were placed on the force transducers to determine how much variability would be introduced by using the system with a wheelchair but while avoiding any effects of an occupant's changing posture and position in the wheelchair. In experiment 3, data were collected with occupied wheelchairs placed on the force transducers to determine the reliability of measurements of weight distribution for each of the four wheels of the occupied wheelchairs. The investigators performed the same activities during each trial.

Calibration weights. Four calibration weights of known mass[dagger] were used for testing (1.13, 2.27, 4.54, and 11.34 kg). A data set was collected by placing each mass on each force transducer 10 times. Several steps were taken to standardize mass placement and force transfer to the transducers, A dot was located at the center of the reference line on each force transducer. A 3.8-cm metal washer was centered on the dot on each force transducer to provide a uniform contact between the masses and force transducers. The force transducers were calibrated to zero by the system with the washers in place. Each time a mass was placed on a transducer, it was centered on the washer with no rotation of the mass compared with the force transducer's reference line.

Unoccupied wheelchairs. Two unoccupied wheelchairs were used. The first wheelchair was a standard wheelchair[double dagger] with urethane tires and casters, swing-away detachable footrests, and full-length detachable armrests. The second wheelchair was an adjustable rigid-frame wheelchair[double dagger] with the rear axles in the most rearward position, pneumatic rear tires, and 12.7-cm (5-in) composite/urethane casters. Each wheelchair was placed on the force transducers three times on 2 separate days for a total of six trials per wheelchair. The larger forefoot force transducers were used under the rear wheels for all testing. The leads from the force transducers were unplugged and reconnected in a crossed fashion before the system was turned on so that the display on the monitor and the information in the printed reports appeared logical. Ramped wooden platforms were fabricated to fit around each force transducer and ease placement of a wheelchair on the transducers (Fig. 2). Several steps were taken to standardize placement of the wheelchairs on the force transducers. One investigator (JA) was responsible for the alignment of each wheelchair for all testing. A dot was placed at the center of each axle on the wheelchairs. A plumb line was used to align each axle over the reference line on the force transducer. There was some tendency for the wheelchairs to roll even though the floor was level. Small rubber chocks, cut from an eraser, were placed in front of and behind each rear wheel to prevent rolling.

Occupied wheelchairs. Criteria for subject participation were (1) that the individual had sustained a cervical spinal cord injury of at least 3 years' duration and (2) that the individual used a manual wheelchair for primary mobility. Three subjects meeting the criteria participated in the study. Subject A had C-8 complete tetraplegia and was 7.5 years postinjury. He used an adjustable wheelchair[sections] and a foam cushion. Subject B had C-6 incomplete tetraplegia and was 6.5 years postinjury. He used an adjustable rigid wheelchair[sections] and an air flotation cushion with minicells. Subject C had C-7 complete tetraplegia and was 10 years postinjury. He used an adjustable folding wheelchair[sections] and a lightweight gel and foam cushion.

The wheelchair was aligned on the force transducers in the manner described for the unoccupied wheelchairs. The computer monitor was positioned out of view of the subjects during testing so that no visual feedback was available from the system, After the wheelchair was positioned, each subject was instructed:

Please place your hands on your hand-rims

as if you were going to propel 

your chair on a level surface. Look

straight ahead and maintain this position

for the next 10 seconds. Start.

The subjects were given a 1-minute rest interval between the taking of five consecutive measurements. The process was repeated on a different day, for a total of 10 trials for each subject.

Data Analysis

The variable monitored was the average vertical force (reported in pounds by the system) measured on each force transducer during each sampling period. This value is not directly measured. It is the product of the "Patient's Weight" value entered prior to testing and the proportion of the total vertical force registered on each force transducer during the sampling period. The value of 100 was entered as the "Patient's Weight" for all testing so that the reported value for each transducer represented a percentage of the total vertical force. This was done for two reasons. The actual weight of a wheelchair and user is usually not readily available in the clinical setting. Additionally, Brubaker[1] discussed the importance of the ratio of weight distributed to the front and rear wheels as a percentage of the total weight. For these reasons, the absolute weight distributed to each wheel was not of interest.

Means were calculated for each series of repeated measures. Standard deviations of the means were calculated to represent the variability of the measures. Data from the rear wheels for each test with a wheelchair were summed to provide a percentage of the total weight distributed to the rear wheels. Intraclass correlation coefficients (ICCs) were calculated[parallel] for each experiment as an index of reliability. The calibration weights and unoccupied wheelchairs were considered to be representative of static loads. The occupied wheelchairs were considered to be representative of the target population of fun-time wheelchair users. Because the same force transducers were used to rate each subject, ICC (2,1) values were calculated as a measure of interjudge agreement.[5]

Results

The data generated when each mass was placed on each force transducer 10 times are represented in Table 1. The values reported are the mean percentage of total vertical force measured by each force transducer for each mass, The standard deviations for these measurements were <0.1%. The standard deviation of measurements of percentage of vertical force exerted by each mass on all four transducers was <0.3%. The ICC for this data set was .99. These percentages were used to calculate a value for each mass for comparison with the actual values and are shown in Table 2. The calculated values were within 0.22 kg of the known values.

[TABULAR DATA 1 & 2 OMITTED]

Table 3 presents the data obtained when unoccupied wheelchairs were repeatedly placed on the force transducers. The standard deviation of repeated measurements of percentage of weight distributed to a wheel was as much as 4.62% on 1 day but was <4% over 2 days. When the percentages of weight distributed to the rear wheels were summed, the standard deviation of six measurements taken on 2 separate days was less than 2%. The ICC for the summed rear-wheel percentages of weight for both wheelchairs on both days was .96.

[TABULAR DATA 3 OMITTED]

The data collected with occupied wheelchairs are shown in Table 4. The means of the five repeated measurements taken on the same day by each force transducer had standard deviations ranging from 0.28% to 4.33%. The greatest difference between day 1 and day 2 means was 3,84% for subject B's right caster. This difference was offset by a 3.34% change in the weight distributed to the left caster. The percentage of weight distributed to the rear wheels for each subject is shown in Figure 3. The standard deviations of the rearwheel weight-distribution measurements for the three subjects over both days ranged from 0.84% to 1.40%, and the ICC for these measurements was .98.

[TABULAR DATA 4 OMITTED]

Discussion

The repeated measurements using calibration weights were taken to determine a baseline level of variability that could be expected from the system without a change in static load on the transducers. As we expected, minimal variability was found. The mass calculated from the test results differed from the known value up to 0.22 kg. Repeated measurements taken with the unoccupied wheelchairs had higher standard deviations than those taken with calibration weights, indicating an increase in variability. This finding was expected because the positioning of the unchanging mass of each wheelchair on the force transducers could not be replicated exactly with each repeated measurement. We expected that the repeated measurements with wheelchairs occupied by subjects would have higher standard deviations, indicative of greater variability, than the unoccupied wheelchairs, but this was not seen. One explanation could be that the lighter unoccupied wheelchairs were positioned more variably on the transducers. We found it more difficult to position the unoccupied wheelchairs because there was more of a tendency of the chair to roll or pivot during the positioning process. An alternate explanation could be that the variability of positioning an unoccupied wheelchair on the force transducers was less significant compared with the greater mass of the occupied wheelchairs.

We expected that the varying posture and positioning of the subjects in their wheelchairs would result in greater variability of repeated measurements, especially between days. The percentage of weight distributed to the rear wheels of the occupied wheelchairs was remarkably consistent. The greatest range among 10 measurements of rear-wheel weight distribution taken on 2 separate days was 4.35%, with a standard deviation of 1.4%, for subject C. The high level of intrarater reliability (one investigator was responsible for positioning the wheelchairs) may be due, in part, to the method used for positioning the wheelchair. The subjects, apparently, also sat in similar postures during the repeated measurements on 2 different days in spite of minimal effort to standardize posture for each measurement. The manner in which the wheelchairs redistribute the forces placed on them may also minimize the effects of small changes in posture on weight distribution.

In Figure 4, approximations of the subjects' rear-axle positions as viewed from the left side of the chair are depicted. Based on rear-axle position, subject C would be expected to have the smallest percentage of weight distributed to the rear wheels. This hypothesis agrees with the findings Fig. 3). Contrary to the findings, however, subject B's more anterior axle placement would be expected to result in the highest rear-wheel weight distribution among the three subjects (Fig. 3). These differences were most likely due to differing wheelchair seat angles and the position of each subject's center of mass. It should be noted that the three subjects were experienced wheelchair users who participated in wheelchair athletics and had adjusted the rear-axle position to their own preference. A considerably smaller percentage of weight was distributed to subject C's left caster during testing (Tab. 4), which could have led to asymmetric wheelchair propulsion characteristics.

Clinicians currently attempt to predict the efficiency and stability of a particular wheelchair for an individual based on factors that can be measured, such as axle position, and on factors that cannot be readily measured, such as the user's center of mass. Use of the Chattecx Balance System" may provide an objective means of evaluating the result of multifactor interactions on manual wheelchair performance. Further investigation is needed to determine the level of interrater reliability and the extent of correlation between functional measures of wheelchair efficiency or stability with the percentage of weight distributed to the rear wheels.

Conclusion

The results of this research indicate that the use of the Chattecx Balance System[TM] for measuring the static weight distribution among the wheels of occupied manual wheelchairs can result in measurements that have excellent reliability. The ability to reliably measure weight distribution will allow systematic investigation of interrelationships that determine manual wheelchair propulsion characteristics and should be valuable in the wheelchair prescription, adjustment, and troubleshooting processes. (*)Chattecx Corp, PO Box 489, Chattanooga, TN 37343-0489. [dagger]Cybex, Div of Lumex Inc, 2100 Smithtown Ave, Ronkonkoma, NY 11779. [double dagger]Invacare Corp, 899 Cleveland St, Elyria, OH 44036. [sections]Quickie Designs Inc, 2842 Business Park Ave, Fresno, CA 93727-1328. [parallel]ICC software developed by Michael J Strube, Department of Psychology, Washington Universit Louis, MO 63130.

References

[1] Brubaker CE. Wheelchair prescription: an analysis of factors that affect mobility and performance. J Rehabil Res Dev. 1986;23(4):19-26. [2] Loane TD, Kirby RL, Static rear stability of conventional and lightweight variable-axle-position wheelchairs. Arch Phys Med Rehabil. 1985;66:174-176. [3] Kirby RL, Kumbhare DA, MacLeod DA, "Bedside" test of static rear stability of occupied wheelchairs, Arch Phys Med Rehabil. 1989;70: 241-244. [4] Lemaire ED, Lamontagne M, Barclay HW, et al. A technique for the determination of center of gravity and rolling resistance for tilt-seat wheelchairs. J Rehabil Res Dev. 1991;28(3): 51-58 [5] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420-427.

JD Tomlinson, PT, is Assistant Professor, Department of Physical Therapy, Beaver College, 450 S Easton Rd, Glenside, PA 19038-3295 (USA). Address all correspondence to Mr Tomlinson.

J Aussprung, PT, is Staff Physical Therapist, Capital Rehabilitation Hospital, Tallahassee, FL 323

H Beatty, PT, is Staff Physical Therapist, Health Care Everywhere, based in Alexandria, VA 22302,

S Patterson, PT, is Staff Physical Therapist, University of Virginia Medical Center, Charlottesvil 22901.

Ms Aussprung, Ms Beatty, and Ms Patterson were students in the Beaver College Physical Therapy Program when this study was completed in partial fulfillment of the requirements for their Master of Science degrees.

This study was approved by the Beaver College Committee on the Protection of Research Subjects.

This article was submitted March 24, 1993, and was accepted December 7, 1993.