Intrasession and intersession reliability of hand-held dynamometer measurements taken on brain-damaged patients.

Intrasession and Intersession Reliability of Hand-held Dynamometer Measurements Taken on Brain-damaged Patients Many methods have been described for assessing motor performance in patients with brain damage (eg, closed head injuries [CHIs] and cerebrovascular accidents [CVAs]). [1-4] For example, to evaluate brain-damaged patients, Fugl-Meyer et al tested reflexes and volitional movements in addition to balance, sensation, range of motion, and coordination. [1] Sheikh and associates assessed motor performance of stroke patients by grading muscle forces, trunk control, and functional tasks. [2] Bobath described the assessment of many motor tasks in her evaluation of patients with hemiplegia. [3] Brunnstrom assessed functional tasks and the patient's ability to move in and out of various postures. [4] Many activities-of-daily-living scales have also been used to assess the motor performance of patients with CVA or CHI. [5]

The use of isokinetic [6] and hand-held [7] dynamometers has been suggested for the assessment of motor performance of hemiplegic patients. Before measurements obtained with dynamometers can be considered useful, they must be demonstrated to be reliable for the population to be measured. [8] For example, the sources of measurement error when measuring forces from limbs of brain-damaged patients are probably different than the sources of error that may be present when measuring forces from limbs of patients with peripheral nerve lesions. Therefore, the degree of reliability may be different for the two populations.

There have been reports that hand-held dynamometers can be used to reliably measure muscle performance in patients with a variety of diagnoses. [7,9] The majority of patients in these studies had brain damage, either a CVA or CHI. Bohannon [7] and Bohannon and Andrews [9] reported high reliability estimates for all measurements taken in their studies. Other researchers have reported lower reliability estimates for hand-held dynamometer measurements taken on other types of patients [10] and on healthy subjects. [11,12] Therefore, we decided to reassess the reliability of measurements taken with a hand-held dynamometer in a manner similar to that used in Bohannon's intratester reliability study. [7]

Bohannon examined the intratester reliability of dynamometer measurements taken on a sample of patients with a variety of neurological diagnoses, the majority of which had diagnoses of CVA. [7] He performed a total of 18 isometric tests on most patients. He took all measurements during a single session with a rest period of 10 to 30 seconds between measurements. The examiner in the study was blind to the result of each test score. Bohannon used the Pearson product-moment correlation coefficient to describe the reliability for the repeated measurements. The reliability values ranged from .84 to .99, with the majority of these values exceeding .96.

In a follow-up report, Bohannon and Andrews examined the intertester reliability for dynamometer measurements taken on a sample of 30 patients with a variety of orthopedic and neurologic diagnoses, including CVA. [9] The two testers participating in the study each measured the isometric force produced by six muscle groups from the upper and lower extremities. They reported that the second examiner typically took his measurements within five minutes of the time the first examiner measured the patient. The Pearson product-moment correlation coefficients used to describe reliability ranged from .88 to .94. In addition, t tests were used to determine whether the means of the measurements of the two raters differed significantly for each of the muscle groups.

We believe there are weaknesses with the reliability studies reported by Bohannon [7] and Bohannon and Andrews. [9] The authors did not address the issue of population-specific reliability in either study. In the intratester reliability study, [7] patients with peripheral neuropathies were included in the sample, and in the intertester reliability study, [9] patients with a variety of nonneurological diagnoses were included. Patients with brain damage who demonstrate motor control problems may demonstrate a very different degree of reliability than patients who have not suffered brain damage. A population-specific sample, therefore, is essential to demonstrate reliability for hand-held dynamometer measurements before these measurements can be used to assess patients who have brain damage.

Bohannon [7] and Bohannon and Andrews [9] did not report whether the patient's position was changed between repeated measurements. The motor performance of patients with central nervous system involvement has traditionally been thought to be affected by numerous temporal and spatial factors. [1,4] Because the investigators did not report whether patients changed positions between repeated measurements, we do not know what effect this issue had on the reliability reported.

Reliability was determined in the studies of Bohannon [7] and Bohannon and Andrews [9] by pooling the data obtained from paretic and nonparetic limbs of patients who had brain damage as well as from limbs of patients without brain damage. Because the motor control problems of patients with brain damage are more manifest in the paretic limbs than in the nonparetic limbs, the reliability of measurements obtained from paretic limbs may be different than the reliability obtained from nonparetic limbs.

Another factor that potentially affects the reliability of hand-held dynamometer measurements is the magnitude of the forces produced by the patient during the test. Bohannon and Andrews [9] and Hosking et al [11] have suggested that measurements of powerful movements are potentially less reliable than nonpowerful movements. Movements of paretic limbs are typically less forceful than movements of nonparetic limbs. This fact may result in a different degree of reliability for measurements obtained from paretic versus nonparetic limbs.

We believe that the statistical analysis method used by Bohannon [7] and Bohannon and Andrews [9] is not appropriate for this type of clinical reliability study. The statistic does not describe the degree of agreement for repeated measurements. The Pearson product-moment correlation, which was used to describe the reliability between paired measurements, is a description of the degree of association, not the degree of agreement, between paired measurements. [8,13] We feel that a more appropriate statistic would have been the intraclass correlation coefficient (ICC), which assesses the degree of agreement between repeated measurements (eg, how closely the first measurement agrees with the second measurement). [13]

Bohannon and Andrews used a t test to determine whether the mean values between two testers were significantly different. [9] A lack of significant differences between means of measurements does not mean that measurements are reliable. In addition, a t test does not estimate the magnitude of the differences associated with repeated measurements.[14]

We believe that the reliability for measurements obtained with hand-held dynamometers on brain-damaged patients must be evaluated further. The purposes of our study were 1) to examine intratester reliability in a manner similar to the procedure used by Bohannon,[7] 2) to determine whether the reliability was similar for measurements obtained from the paretic versus the nonparetic limbs of patients with brain damage, and 3) to examine the degree of reliability of hand-held dynamometer measurements over a more clinically meaningful time interval. This time interval had to be long enough so that the stability of the measurement could be demonstrated. As Rothstein has noted, if a measurement varies greatly over a brief period of time, perhaps the phenomena being measured cannot be measured reliably.[8] The time interval we chose also had to be short enough so that treatment or recovery would not lead to changes in the patient's ability to produce muscle forces.

Method

Our study was divided into two parts. Part 1 was similar to Bohannon's study[7] in which repeated dynamometer measurements were taken by the same examiner on a sample of brain-damaged patients within a single session. Part 2 consisted of two testing sessions that were 48 hours apart.

Subjects

Thirty subjects (14 men, 16 women) participated in Part 1 of the study, and 15 subjects (9 men, 6 women) participated in Part 2. One subject participated in both parts of the study. All subjects were patients referred to the Department of Physical Therapy, Medical College of Virginia Hospital, Richmond, Va. The criteria for admission to the study were that the subject have brain damage and be able to follow the simple instructions given during the testing procedure.

The mean age of the subjects in Part 1 was 54.6 years (s = 18.1), with a range of 23 to 85 years. A total of 21 subjects in Part 1 had a diagnosis of CVA, 8 subjects had a diagnosis of CHI, and 1 subject had suffered a stab wound to the head. The mean amount of time since onset of brain damage for the subjects in Part 1 was 52 days for those with CVA and 142 days for those with CHI. Eight subjects in Part 1 were paretic on only their right side, and 17 were paretic on only their left side. Five subjects were bilaterally paretic (Tab. 1). The side or sides of paresis was determined by the therapists treating the patients.

The mean age of the subjects in Part 2 was 53.9 years (s = 17.5), with a range of 18 to 76 years. A total of 10 subjects in Part 2 had a diagnosis of CVA, and 5 subjects had a diagnosis of CHI. The mean amount of time since onset of brain damage for the subjects in Part 2 was 49 days for those with CVA and 96 days for those with CHI. Four subjects in Part 2 were paretic on only their right side, and 8 were paretic on only their left side. Three subjects were bilaterally paretic (Tab. 1).

Procedure

The procedure used in both parts of this study was based on the method originally described by Bohannon.[7] We performed 10 isometric tests instead of the 18 used by Bohannon. Subjects read and signed a consent form prior to participation in the study. A hand-held dynamometer(*1) was used to measure the forces produced during isometric contractions of muscles that would normally produce the following movements: wrist flexion, wrist extension, elbow flexion, elbow extension, shoulder internal (medial) rotation, shoulder external (lateral) rotation, hip flexion, ankle dorsiflexion, knee flexion, and knee extension. All tests except thos for knee flexion and knee extension were conducted with the subjects seated with their legs dangling over the edge of a mat table. Subject position, dynamometer placement, and method of subject stabilization were as described by Bohannon.[7] Some subjects reported discomfort in the area of the interface between the dynamometer and the skin overlying the distal tibia during knee extensor tests. We used a towel between the subject's skin and dynamometer during these tests. Bohannon did not report whether he calibrated the dynamometer he used to take measurements. Therefore, we did not check the calibration of the dynamometer and reported all values in relative units. The tests performed in this study can be considered similar to isometric manual muscle tests except that a dynamometer was used to quantify the forces produced by the subject.

Three measurements were made of each isometric test. A recorder (SDF or MLW) read the force produced by the subject to the nearest one-half kilogram, as marked on the dynamometer force gauge, and recorded the measurements. Between each measurement, the only examiner in the study (DLR) gently moved the limb segment being tested through the available pain-free ROM. We felt this procedure was necessary to more closely replicate the clinical situation, in which patients typically change positions between serial measurements. Bohannon did not report in his study[7] whether subjects' limbs were moved between measurements.

Between measurements, there was a rest period of 10 to 30 seconds. Each resisted movement was held by the subject to 4 to 6 seconds. The examiner palpated the muscles being tested, and when he determined that no contraction occurred, data were not collected. When the force produced by the subject exceeded the scale of the force gauge, the results of the test were not included in the data analysis.

For Part 2 of the study, the same method was used except that a second test session was conducted approximately 48 hours after the first series of measurements were taken. The second measurement session was conducted at the same time of day as the first measurement session.

Data Analysis

Intraclass correlation coefficients (formula 1,1)[15] were calculated to describe the degree of intratester agreement for the measurements taken in this study. Although we have argued that the ICC is the best statistic to use in this type of study, we also calculated Pearson product-moment correlation coefficients (r) to compare or results with those of Bohannon.[7] In addition, coefficients of determination ([r.sup.2]) were calculated to provide a description of the percentage of common variance between repeated measurements, which allows a more understandable comparison between the Pearson product-moment correlation coefficients and the ICCs.

In Part 1, calculations of the ICCs were made by comparing all three measurements for each muscle group taken in the test session. Pearson product-moment correlations and coefficients of determination were calculated by comparing the first and second measurements, the second and third measurements, and the first and third measurements. Correlations (ICC, [r.sup.2], r) were calculated for the force measurements obtained from the paretic limbs and the nonparetic limbs. In addition, correlations were calculated from the combined data from both the paretic and nonparetic limbs.

For Part 2, correlations (ICC, [r.sup.2], r) were calculated by comparing the first measurements of the first session to the first measurements of the second session. Correlations were calculated for the force measurements obtained from the paretic limbs, the nonparetic limbs, and the combined data from both limbs. To assess whether reliability could be improved by taking the means of multiple measurements, correlations were also calculated for the average of the three measurements for each muscle group from each test session.

Results

Part 1

The ICC values for the repeated measurements obtained from the paretic limbs during a single session ranged from .93 to .98. The ICC values for the measurements obtained from the nonparetic limbs ranged from .88 to .98 (Tab. 2). The ICC values for the measurements obtained from the combined data from both limbs ranged from .90 to .98 (Tab. 3). The coefficients of determination and Pearson product-moment correlation coefficients are also reported in Tables 2 and 3.

Part 2

The ICC values for the first measurements obtained from the paretic limbs during two sessions ranged from .90 to .98 (Tab. 4). The ICC values for the average of the three measurements obtained from the paretic limbs ranged from .87 to .99 (Tab. 5). Regardless of whether the first measurements or the averages were correlated, the majority of the ICCs for measurements taken from the paretic limbs were .95 or above.

The ICC values for the first measurements obtained from the nonparetic limbs ranged from .31 to .93 (Tab. 4). The ICC values for the average of the three measurements obtained from the nonparetic limbs ranged from .28 to .93 (Tab. 5). Regardless of whether the first measurements or the averages were correlated, the majority of the ICCs for measurements taken from the nonparetic limbs were .78 or below.

The ICC values for the first measurements obtained by combining the data from all the limbs ranged from .79 to .95 (Tab. 4). The ICC values for the average of the three measurements obtained from the combined data ranged from .81 to .97 (Tab. 5). The coefficients of determination and Pearson product-moment correlation coefficients are also reported in Tables 4 and 5.

Discussion

Part 1

When data from a single session were compared, we found a high degree of reliability for force measurements of resisted movements of paretic and nonparetic limbs and for pooled data from all limbs. However, we observed that the reliability values for force measurements taken from the nonparetic limbs in Part 1 (within a session) of our study were slightly lower, in general, than the values for the measurements taken from the paretic limbs. Bohannon also reported high levels of reliability in his single-session intratester reliability study when he examined pooled data from the paretic and nonparetic limbs. [7]

Part 2

When the time interval between repeated measurements was increased to two days, a high degree of reliability was found only for the force measurements taken from the subject's paretic limbs and for the pooled data. Reliability estimates were lower, sometimes drastically, for the measurements taken from the nonparetic limbs.

This finding of greater reliability for measurements taken from the paretic side than from the nonparetic side was not anticipated. Because motor control problems are typically more evident on the paretic sides of brain-damaged patients, we initially thought that the reliability of measurements from the paretic limbs would have been lower than for the nonparetic limbs. The design of our study did not allow us to determine why greater reliability existed for measurements of the paretic limbs.

One possible explanation for the differences in reliability between measurements taken from paretic and nonparetic limbs of brain-damaged patients may be motor control differences between the two sides. Patients with CVA and CHI are forced, because of their brain damage, into stereotypical movements on their paretic side. [4] As a result, repeated measurements of muscle forces would tend to be similar when tested in the same way. These measurements may be in contrast to measurements taken from the nonparetic side. Forces generated by nonparetic limbs may be more free to vary than forces generated by paretic limbs, and as a result, there may be less agreement between repeated measurements.

Other authors have hypothesized that the reliability of dynamometer measurements may be related to the magnitude of the forces being measured. [9,11] They observed that when greater forces are produced by the patient, more stabilization by the therapist is required. Therefore, the more stabilization a test requires, the more difficult that test is to perform. They suggest that such tests are likely to be less reliable than tests where very little force is being measured.

To determine whether a relationship existed between the forces we measured and the reliability of those measurements, an a posteriori analysis was conducted. Using data from Part 2, we correlated the mean force obtained for each test of the nonparetic limb (Tab. 6) with the corresponding ICC value for that test. The Pearson product-moment correlation was. 10, which suggests there was little relationship between the magnitude of force level and the degree of reliability.

Our study had some limitations that potentially limit the generalizability of our results. The number of subjects in our sample was small. We used only one tester to take the measurements. We used only one type of hand-held dynamometer; therefore, the results may not be generalizable to measurements obtained with other dynamometers. However, our method and sample were comparable to those already described in the literature.

The results of our study and those of Bohannon [7] and Bohannon and Andrews [9] appear to indicate that hand-held dynamometer measurements obtained from brain-damaged patients are reliable. However, reliability is not the only issue that must be examined before measurements can be considered clinically useful. If therapists use hand-held dynamometers to measure muscle performance on brain-damaged patients, they must consider whether the measurements provide meaningful information. For example, Bohannon and Smith have suggested that hand-held dynamometer measurements can be used to make clinical decisions relating to rehabilitation potential and the planning of treatment. [16] They appear to believe that the rehabilitation potential of hemiparetic patients can be predicted from force measurements obtained at the time the patient is admitted to a rehabilitation program. In addition, they argue that treatment planning could be based on the force measurements. They base these opinions on "force deficits" they observed at initial assessment and those they observed at discharge. A force deficit was defined as the difference in force measurements between the nonparetic limb measured at admission and the force of the resisted movement of the paretic limb. The deficits were expressed as percentages by dividing the differences in forces by the force values for the nonparetic limbs at the time of admission and then multiplying by 100. Therefore, all force measurements were expressed in terms of those of the nonparetic limb at the time of admission.

Our results indicate that hand-held dynamometer measurements of the nonparetic extremity have poor to moderate reliability. Based on the poor reliability estimates for the nonparetic limbs, measurements from the nonparetic side should not be used for comparisons with the paretic side. Therefore, we question the validity of the inferences relating to rehabilitation potential and treatment planning as suggested by Bohannon and Smith [16] because they used measurements from nonparetic limbs to examine these inferences.

Further research is needed on the validity of hand-held dynamometer measurements taken on brain-damaged patients. We have noted that measurements taken from paretic limbs appear to be highly reliable. Studies examining the usefulness of hand-held dynamometer measurements taken on paretic limbs have the potential for clarifying the role of instrumented muscle testing in the brain-damaged population.

Conclusion

Force measurements taken with a hand-held dynamometer on brain-damaged patients are highly reliable when repeated measurements are taken within a very short time interval. When the repeated measurements are taken after a more clinically meaningfuly time interval (ie, two days), the reliability remains high for measurements taken on the paretic limbs. However, when the nonparetic limbs are measured over a period of two days, the reliability varies depending on the movement measured, with reliability being poor for most measurements. Even though some of the measurements taken in this study appear to be highly reliable, their usefulness must be questioned because measurements of muscle performance of brain-damaged patients currently lack a theoretical basis.

Further research should focus on examining the validity of hand-held dynamometry for measuring muscle performance of brain-damaged patients (ie, determine whether clinical inferences can be made from these measurements). These studies must be performed before clinicians use these measurements for decision making.

Acknowledgments

We thank Thomas Mayhew, PT; Robert Lamb, PhD, Pt; Ann VanSant, PhD, PT; and Denise Akromas, PT, for their assistance in the preparation of this manuscript. We also thank the Rehabilitation Team in the Physical Therapy Department at Medical College of Virginia Hospital for their assistance during data collection.

(*1) Spark Instruments and Academics, Inc, PO Box 5123, Coralville, IA 52241.

[1] Fugl-Meyer AR, Jaasko L, Leyman I, et al: The post-stroke hemiplegic patient: I. A method for evaluation of physical performance. Scand J Rehabil Med 7:13-31, 1975

[2] Sheikh K, Smith DS, Meade TW, et al: Assessment of motor function in studies of chronic disability. Rheumatol Rehabil 19:83-90, 1980

[3] Bobath B: Adult Hemiplegia: Evaluation and Treatment, rev ed 2. London, England, William Heinemann Medical Books Ltd, 1978, pp 16-22

[4] Brunnstrom S: Movement Therapy in Hemiplegia. Hagerstown, MD, Harper & Row, Publishers Inc, 1970, pp 34-56

[5] Wade DT, Hewer RL, Skilbeck CE, et al: Stroke: A Critical Approach to Diagnosis, Treatment, and Management. Chicago, IL, Year Book Medical Publishers Inc, 1985, pp 151-156

[6] Watkins MP, Harris BA, Kozlowski BA: Isokinetic testing in patients with hemiparesis: A pilot study. Phys Ther 64:184-189, 1984

[7] Bohannon RW: Test-retest reliability of hand-held dynamometry during a single session of strength assessment. Phys Ther 66:206-209, 1986

[8] Rothstein JM: Measurement in clinical practice: Theory and application. In Rothstein JM (ed): Measurement in Physical Therapy: Clinics in Physical Therapy. New York, NY, Churchill Livingstone Inc, 1985, vol 7, pp 8-9

[9] Bohannon RW, Andrews AW: Interrater reliability of hand-held dynamometry. Phys Ther 67:931-933, 1987

[10] Wiles CM, Karni Y: The measurement of muscle strength in patients with peripheral neuromuscular disorders. J Neurol Neurosurg Psychiatry 46:1006-1013, 1983

[11] Hosking GP, Bhat US, Dubowitz V, et al: Measurements of muscle strength and performance in children with normal and diseased muscle. Arch Dis Child 51:957-963, 1976

[12] Clarke HH: Comparison of instruments for recording muscle strength. Research Quarterly 25:398-411, 1954

[13] Bartko JJ, Carpenter WT: On the methods and theory of reliability. J Nerv Ment Dis 163:307-317, 1976

[14] Linton M, Gallo PS Jr: The Practical Statistician: Simplified Handbook of Statistics. Monterey, CA, Brooks/Cole Publishing Co, 1975, pp 276-279

[15] Shrout PE, Fleiss JL: Intraclass correlations: Uses in assessing rater reliability. Psychol Bull 86:420-428, 1979

[16] Bohannon RW, Smith MB: Upper extremity strength deficits in hemiplegic stroke patients: Relationship between admission and discharge assessment and time since onset. Arch Phys Med Rehabil 68:155-157, 1987

D Riddle, MS, PT, is Assistant Professor, Department of Physical Therapy, School of Allied Health Professions, Medical College of Virginia, Commonwealth University, PO Box 224, MCV Station, Richmond, VA 23298-0224 (USA).

S. Finucane, BS, PT, is a graduate student, Department of Physical Therapy, Medical College of Virginia.

J Rothstein, PhD, PT, is Associate Professor, Department of Physical Therapy, Medical College of Virginia.

M Walker, MS, is Assistant Professor, Program in Physical Therapy, College of Health Sciences, School of Community Health and Physical Therapy, Old Dominion University, Norfolk, VA 23508. She was Instructor, Program in Physical Therapy, College of Health Sciences, School of Community Health and Physical Therapy, Old Dominion University, when this study was conducted.

This article was submitted October 30, 1987; was with the authors for revision for eight weeks; and was accepted March 25, 1988.