Physical Performance Tasks: What Are the Underlying Constructs?

44

Physical Performance Tasks: What Are the Underlying Constructs? Diane M. Novy, PhD, Maureen J. Simmonds, PT, PhD, C. Ellen Lee, PT, MS ABSTRACT. Novy DM, Simmonds MJ, Lee CE. Physical performance tasks: What are the underlying constructs? Arch Phys Med Rehabil 2002;83:44-7. Objectives: To investigate the structural validity of a battery of physical performance tasks and to investigate the construct validity of the resulting factors. Design: A measurement study. Setting: A large, private orthopedic clinic and a physical therapy clinic in an urban area. Participants: One hundred three consecutive adult patients with low back pain who were referred for physical therapy assessment. Interventions: Not applicable. Main Outcome Measures: The physical performance battery was composed of the 50-foot speed walk, 5-minute walk, repeated trunk flexion, sit to stand, loaded reach, rollover tasks, and Sorensen upper-body lift. Participants also completed 5 self-report measures. Results: Two correlated (.74) factors, speed and coordination and endurance and strength, were derived from the physical performance tasks. The Sorenson upper-body lift task was the only indicator that was not useful in defining the factors. Both factors had statistically significant correlations with measures of physical disability, lack of self-efficacy, and negative affect. Both factors had a trivial correlation with a numeric rating of pain intensity. Conclusions: There is support for 2 meaningful empiric groupings (ie, the derived factors) of the performance tasks. Pain intensity had a trivial overlap with speed-coordination and endurance-strength factors. Key Words: Low back pain; Physical endurance; Rehabilitation. © 2002 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation OW BACK PAIN (LBP) is a common, complex, costly, L and controversial problem. Traditionally, standard clinical assessments of LBP have focused on the degree of impairment in the lower back (eg, range of motion, muscle strength as they relate specifically to the back). The present trend is to comple-

From the Departments of Anesthesiology and Psychiatry and Behavioral Sciences, University of Texas-Houston Health Science Center, Houston, TX (Novy); and the Department of Physical Therapy, Texas Woman’s University, Houston, TX (Simmonds, Lee). Accepted March 7, 2001. Supported in part by Texas Woman’s University Research Enhancement Award Grant (grant no. 10-0131382) and the National Institutes of Health Extramural Associates Research Development Award research pilot project (grant no. 1997-8). 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. Reprint requests to Diane M. Novy, PhD, Dept of Anesthesiology, University of Texas-Houston Medical School, 6431 Fannin, MSB 5.020, Houston, TX 77030, e-mail: [email protected]. 0003-9993/02/8301-6621$35.00/0 doi:10.1053/apmr.2002.27397

Arch Phys Med Rehabil Vol 83, January 2002

ment impairment measures with functionally based measures that address more globally a person’s physical limitations. The present study addresses unresolved measurement issues on a battery of physical performance tasks that are examples of functionally based measures. The current status of physical performance tasks has evolved from studies of patients with LBP and normal controls. In the first studies, Simmonds et al1-3 investigated various performance tasks to represent the domain of physical performance. Based on solid interrater agreement, test-rest reliabilities, and convergent and discriminant validity, they recommend using the 50-foot speed walk, 5-minute walk, repeated trunk flexion, sit to stand, loaded reach, and Sorensen upper-body lift tasks as indicators of overall physical performance. In a subsequent study,4 these researchers reported preliminary data to suggest that 2 related constructs underlie the physical performance tasks. Specifically, they used a discriminant function analysis to investigate gender differences on the performance tasks. This analysis revealed constructs pertaining to endurance and strength and to velocity and coordination. The velocity-coordination construct was defined by 5-minute distance walk and loaded reach, whereas the endurance-strength construct was defined by 50-foot speed walk, trunk flexion, and sit to stand tasks.4 The Sorensen upper-body lift task appeared to be a weak indicator of these constructs. However, confirmation or disconfirmation of the use of the upper-body lift task needs to be investigated further. In a recent effort to expand the performance task battery, Simmonds’s1-3 team piloted another a rollover task on a sample of patients with LBP. A needed next step is to assess the use of the rollover task in conjunction with the rest of the established performance battery. The purpose of the present study was to identify the performance tasks that contribute to the overall structural validity of the performance battery, giving particular attention to the questionable upper-body lift task and the previously unresearched rollover task. Related to this, it is important to investigate common factors that underlie the different performance tasks. To gain a clearer understanding of the constructs, it also is important to investigate patterns of correlation between the constructs and external variables such as depression, pain intensity, physical limitation, and self-efficacy related to the performance of certain functional tasks. METHODS Participants The participants were 103 consecutive adult patients with LBP referred for physical therapy assessment from a large, private orthopedic clinic and a physical therapy clinic in an urban area. The physical performance battery was incorporated as part of the routine physical therapy clinical intake examination. The participants ranged in age from 19 to 82 years (average age ⫾ standard deviation [SD], 45 ⫾ 12.88yr). Sixtyone participants were women. Fifteen percent of the participants had a previous back surgery. Other clinical characteris-

PHYSICAL PERFORMANCE TASKS, Novy Table 1: Clinical Characteristics of the Sample (N ⴝ 103) Characteristics

Mean ⫾ SD

Duration of pain (mo) NRS-Pain NRS-Affect BDI-SF SES* RMDQ 50-foot walk (s) 5-minute walk Repeated trunk flexion (s) Repeated sit to stand (s) Loaded reach (cm) Rollover (s) Sorensen fatigue (s)

89.97 ⫾ 141.85 4.40 ⫾ 2.59 5.54 ⫾ 2.60 4.82 ⫾ 4.91 20.25 ⫾ 11.14 10.02 ⫾ 5.86 10.97 ⫾ 3.47 398.78 ⫾ 109.19 11.47 ⫾ 3.94 13.00 ⫾ 6.29 53.78 ⫾ 19.28 14.93 ⫾ 8.45 34.68 ⫾ 47.07

* Assesses degree of worry related to performing certain functional tasks.

tics of the sample, including data on the performance tasks and external self-report measures, are shown in table 1. Procedure All procedures followed the guidelines for ethical standards at the University of Texas-Houston Health Science Center and Texas Woman’s University. Data on the participants were collected from 1997 to 1999. After the subjects gave written consent to participate in the study, they completed several self-report questionnaires and then completed the performance tests. The physical performance battery was administered by 1 of 4 licensed physical therapists in a procedure that required approximately 20 minutes. Measures In addition to the 7 physical performance tasks discussed later, we also included several self-report measures so that construct validity of the resulting factors could be investigated. Physical performance tasks. With the exception of the Sorensen upper-body lift task, the performance tasks were administered in counterbalanced order. The upper-body lift task was always performed last because it is the most physically challenging task. Fifty-foot walk. This task requires the subject to walk along a 25-foot walkway, turn around, and walk back to the start as fast as possible. The total time it takes to perform this task is the resulting score. Five-minute walk. This task requires the subject to walk as far as possible in 5 minutes. The distance walked is the resulting score. Repeated trunk flexion. This task requires the subject to bend forward (as if touching his/her toes) and, as fast as tolerated, return to upright standing 5 times. The task is repeated after a brief pause and the average time of the 2 tasks is the resulting score. Repeated sit to stand. This task uses a standard armless chair and requires the subject to stand from a sitting position and, as fast as possible, to return to the seated position 5 times. The task is repeated after a brief pause and the average time of the 2 tasks is the resulting score. Loaded reach. To perform this task, a light wooden bar with sandbag weights of 5% of the person’s body weight (not exceeding 4.5kg) is held in both hands at shoulder height close to the body. Subjects are required to reach forward as far as possible. The distance reached is the resulting score.

45

Rollover. This task requires the subject to lie supine and then to roll over a full 360° as fast as possible. After a brief pause, they roll 360° in the opposite direction. The order of direction is randomized. Each direction of rolling is timed separately and the average time taken is the resulting score. Sorensen upper-body lift. This task requires the subject to lay prone on a standard treatment table, positioned with the anterior sacroiliac spines at the edge of the table. Straps are placed across the thighs and calves for stabilization. With arms straight at the sides, the person must lift the upper body so that it is in a horizontal plane with the table. That position is to be held as long as possible. Time until fatigue is the resulting score. Self-Report Measures Subjects completed the 5 self-report surveys listed below. Roland-Morris Disability Questionnaire. The RolandMorris Disability Questionnaire5,6 (RMDQ) is a 24-item measure of current activity limitation from LBP. The total RMDQ score is a sum of activities that are limited, with higher total scores representing greater activity limitation. Self-Efficacy Scale. The Self-Efficacy Scale7 (SES) is an 8-item measure that probes how confident a person is that he/she can perform certain functional tasks. A 7-point Likert scale is the response scale. The total SES score is a sum of those ratings, with higher total scores suggesting greater worry in regard to performing the listed tasks. Beck Depression Inventory–Short Form. The Beck Depression Inventory–Short Form8 (BDI-SF) is a 13-item measure of depressive symptoms. Items are rated on a 0 to 3 scale according to how a respondent felt over the past week. The total BDI-SF score is a sum of the ratings, with higher scores suggesting more depression. Numeric rating scale of pain affect. A numeric rating scale of pain affect (NRS-Affect) was used to assess current pain unpleasantness. Anchors for the scale are 0 (not unpleasant) and 10 (extremely unpleasant). Numeric rating scale of pain intensity. A numeric rating scale of pain intensity (NRS-Pain) was used to assess current pain intensity. Anchors for the scale are 0 (no pain) and 10 (worst pain imaginable). Data Analyses A preliminary data analysis using the intercorrelation coefficient procedure was run to determine the previously untested reliability of the rollover task. Test-retest reliability was .89 and intertester reliability was .81. Although data from a previous discriminant function analysis suggested a potentially meaningful factor structure for the performance tasks,4 those findings are considered too preliminary to support testing a specific factor model with confirmatory factor analytic procedures. In the present study, exploratory factor analysis was used to uncover the underlying constructs from the correlation matrix of the 50-foot speed walk, 5-minute walk, repeated trunk flexion, sit to stand, loaded reach, Sorenson upper-body lift, and rollover performance tasks. Principal axis was the common factor extraction method and oblimin was the rotation method by which we derived correlated constructs. RESULTS A scree plot and inspection of the resulting factors suggested the retention of 2 factors. The first factor had an eigenvalue of 4.26 and accounted for 60.88% of the variance in the data set. The second factor had an eigenvalue of .95 and accounted for 13.59% of the remaining variance in the data set. Arch Phys Med Rehabil Vol 83, January 2002

46

PHYSICAL PERFORMANCE TASKS, Novy Table 2: Factor Pattern Coefficients

Table 4: Correlations of Factors With Self-Report Measures †

Performance Task

Factor 1*

Factor 2

50-foot walk 5-minute walk Repeated flexion Sit to stand Loaded reach Rollover Sorenson

.72 ⫺.05 .92 .91 .01 .86 ⫺.30

⫺.26 .95 .04 .07 .60 .02 .04

* Speed and coordination. Endurance and strength.

†

The resulting factors had a substantial amount of overlap (r ⫽ .74). The factor pattern (standardized regression coefficients) and factor structure (correlations coefficients between the individual performance tasks and the factors) matrices appear in tables 2 and 3, respectively. Although pattern and structure coefficients are used for interpreting correlated factors, pattern coefficients are easier to interpret because they ignore the correlation between factors. Salient indicators for the first factor were the 50-foot speed walk, repeated trunk flexion, sit to stand, and rollover tasks. These 4 indicators are associated with speed, coordination, and the ability to withstand constantly and quickly changing spinal loads. Salient indicators for the second factor were the 5-minute distance walk and loaded reach tasks. These 2 indicators are associated with strength, endurance, and the ability to withstand static or relatively small changes in spinal load. Unlike the salient indicators, the Sorensen upper-body lift task contributed a trivial amount of variance to the speed-coordination factor and even less variance to the endurance-strength factors. Factor scores for the 2 factors were derived with the regression method. The resulting factor scores then were correlated with the external, self-report measures to define further the factors and establish construct validity. These correlations are in table 4. As expected, both physical performance factors had negative correlations with the self-report measures of affective distress, pain intensity, physical disability, and worry about performing certain functional tasks. Each factor was moderately correlated with perceived physical disability as measured by the RMDQ and with worry about performing functional tasks as measured by the SES. Each factor correlated to a lesser degree with negative affect as measured by the BDI-SF. Factor 2, endurance and strength, correlated with the numeric rating of affect; however, factor 1, speed and coordination, had only a trivial correlation with the numeric rating of affect. Both factors correlated trivially with the numeric rating pain intensity.

Table 3: Factor Structure Coefficients Performance Task

Factor 1*

Factor 2†

50-foot walk 5-minute walk Repeated trunk flexion Sit to stand Loaded reach Rollover Sorenson

.90 ⫺.72 .89 .86 ⫺.41 .84 ⫺.32

⫺.77 .98 ⫺.60 ⫺.57 .59 ⫺.58 .25

* Speed and coordination. † Endurance and strength.

Arch Phys Med Rehabil Vol 83, January 2002

Self-Report Measures

NRS-Affect NRS-Pain RMDQ SES BDI-SF

Factor 1*

Factor 2†

⫺.17 ⫺.16 ⫺.58 ⫺.45 ⫺.31

⫺.34 ⫺.21 ⫺.48 ⫺.49 ⫺.24

NOTE. Correlations of absolute value of ⱖ .31 are significant at .01. Correlations of absolute value of ⱖ .21 are significant at .05. * Speed and coordination. † Endurance and strength.

DISCUSSION Although considerable psychometric support existed for the 50-foot speed walk, 5-minute walk, repeated trunk flexion, sit to stand, loaded reach, and upper body lift as indicators of physical performance, there was limited information on the underlying factor structure for these particular tasks. Indeed, some preliminary data cast doubt on the contribution of the Sorensen upper-body lift task to the overall structural validity of the physical performance battery. On the other hand, a rollover performance task may be a useful indicator of physical performance. However, its unique contribution as an indicator with the rest of the performance task battery has not been tested. Derivation of structural validity is a potentially useful next step in this program of research for 2 reasons. First, this information will support or refute the use of the Sorensen upper-body lift and rollover tasks as indicators in the battery. Second, information at the construct level may help guide physical therapy treatment protocols. Information at the construct level also may be useful in communicating with nonphysical therapists on the treatment team and with patients who might find information about the individual tasks too specific and technical. For those reasons, the present study investigated the structural validity of the performance tasks. Specifically, exploratory common factor analysis was used to uncover the structural validity of the 50-foot speed walk, 5-minute walk, repeated trunk flexion, sit to stand, loaded reach, upper-body lift, and rollover performance tasks. Two factors were retained that correlated .74 with each other. Factor 1 was named speed and coordination. This factor was identified by the 50-foot speed walk, repeated trunk flexion, sit to stand, and rollover. The rollover task had a .86 loading on this factor and appears to be a strong indicator of speed and coordination. Factor 2 was named endurance and strength. The distance walked in 5 minutes and the loaded reach task identified this factor. In contrast to the other tasks tested here, the Sorensen upper-body lift task was not associated with salient loadings on either factor. Hence, its overall usefulness in conjunction with the rest of the performance task battery is questionable relative to the other performance tasks. The 2 factors, speed and coordination and endurance and strength, may be useful in guiding physical therapy treatment. When scores of physical performance tasks indicate patients are compromised in speed and coordination of movement, clinicians could design exercise protocols to target the underlying factor. Likewise, patients with poor performance on indicators of endurance and strength may benefit from a training protocol that targets and enhances these deficits. In a study by de Looze et al,9 it was shown that reduced strength compromises the capability of the spine to withstand the mechan-

PHYSICAL PERFORMANCE TASKS, Novy

ical load. Therefore, patients may have better endurance in withstanding spinal loading in both the 5-minute walk and loaded reach tasks after the training. Given the strong correlation between the factors, any treatment designed to improve functioning in 1 factor will likely impact the other factor as well. The strong correlation between the factors also highlights the relevance of overall physical performance. To understand the meaning of the related factors better, we explored how the 2 factors correlated with external, self-report measures. As expected, all correlations with measures of disability, worry about performing functional tasks, and negative affect and pain were negative. Moderately high correlations were found for the factors with perceived physical disability and with worry about performing functional tasks. Each factor correlated to a lesser degree with negative affect. The endurance-strength factor correlated with the numeric rating of affect. The speed-coordination factor had only a trivial correlation with the numeric rating of affect. Both factors had a trivial amount of overlap with pain intensity. Considering these correlations together, it appears that there are 3 overlapping, limiting aspects of physical performance: perceived physical disability, lack of self-efficacy, and negative affect. Although the findings from the present study are potentially useful for improving understanding of the factors and possibly LBP treatment, several design limitations should be considered. First, this study examined the factors among LBP patients who were referred from secondary treatment centers (ie, outpatient physical therapy clinic and orthopedic surgeon clinic); therefore, the results may be generalizable only to this population. Second, a specific battery of physical performance tasks was included. Still to be tested is whether this particular battery is optimal or whether other physical performance tasks may add to the definition of factors. Directions for future research include investigating the structural stability of the factors derived with the present sample of patients. If findings are corroborated, confirmatory factor analytic procedures could be used to test specific factor structures. It also will be important to see the extent to which the factors are invariant across other samples (eg, tertiary pain treatment centers) and other physical conditions (eg, fibromyalgia). An-

47

other future research direction will be to include other external measures in subsequent investigations of construct validity. CONCLUSIONS Two correlated factors, speed and coordination and endurance and strength, were derived from a battery of physical performance tasks. Before this derivation of factors, communication was possible on only 2 levels: scores on the individual tasks and an overall assessment of physical performance. Now there is support for 2 meaningful groupings (ie, the derived factors) that underlie the individual tasks. Finally, both factors overlapped with perceived physical disability, lack of selfefficacy, and negative affect. Because both factors had trivial overlap with pain intensity, it does not appear to be a major limiting aspect of physical performance. References 1. Simmonds MJ, Olson SL, Novy DM, et al. Classification of subjects with and without low back pain using novel tasks of physical performance. In: Proceedings of the 16th Annual Scientific Meeting of the American Pain Society; 1997 Oct 23-26; New Orleans (LA). 2. Simmonds MJ, Olson SL, Jones S, et al. Psychometric characteristics and clinical usefulness of physical performance tests in patients with low back pain. Spine 1998;23:2412-21. 3. Simmonds MJ, Olson S, Novy DM, Jones SC. Disability prediction in patients with back pain using performance based models. In: Proceedings of the North American Spine Society/American Pain Society Meeting; 1998 April; Charleston (SC). 4. Novy DM, Simmonds MJ, Olson SL, Lee CE, Jones SC. Physical performance: differences in men and women with and without low back pain. Arch Phys Med Rehabil 1999;80:195-8. 5. Roland M, Morris R. A study of the natural history of back pain. Part I: development of a reliable and sensitive measure of disability in low-back pain. Spine 1983;8:141-4. 6. Roland M, Morris R. A study of the natural history of low-back pain. Part II: development of guidelines for trials of treatment in primary care. Spine 1983;8:145-50. 7. Bernstein IH, Jaremko ME, Hinkley BS. On the utility of the SCL-90-R with low-back pain patients. Spine 1994;19:42-8. 8. Reynolds WM, Gould JW. A psychometric investigation of the standard and short form Beck Depression Inventory. J Consult Clin Psychol 1981;49:306-7. 9. de Looze MP, Zinzen E, Caboor D, van Roy P, Clarijs JP. Muscle strength, task performance and low back load in nurses. Ergonomics 1998;41:1095-104.

Arch Phys Med Rehabil Vol 83, January 2002