Contribution of Psychosocial and Mechanical Variables to Physical Performance Measures in Knee Osteoarthritis

Method

The data reported here are part of a larger research project in which subjects made 2 visits of approximately 2 hours each, 1 week apart. All measurements reported here were obtained on the second visit. Data from the first visit, a kinematic and kinetic gait analysis, will be reported elsewhere.

Results

Table 1 provides the means, standard deviations, and ranges of values obtained for the physical performance measures and the psychosocial and mechanical variables. One subject did not complete the strength testing because of a medical event related to asthma. A repeated-measures analysis of variance revealed that fatigue was not a factor over the 5 strength trials of knee extension and flexion (P=.61 and P=.76, respectively).

The subjects had a mean BMI of 28.6 kg/m² (SD=5.1), indicating that, as a group, the subjects were overweight. The mean ratio of QUAD to HAM muscle strength was 0.52 (SD=0.19). The responses on the CES-D Scale and STAI measures indicated that the levels of depression and anxiety found in the general population were lower than those found in people with psychiatric disorders.

Correlation coefficients (r) for the FSE and all physical performance measures ranged between absolute values of .68 and .72 (P<.001) (Tab. 2). The STAI data were not significantly related to data obtained for the performance tasks, and the correlations (r) between the CES-D Scale data and data obtained for the performance tasks ranged between absolute values of .33 and .37. Correlation coefficients (r) for the mechanical variables (BMI, QUAD, HAM) and the physical performance measures ranged between absolute values of .35 and .52.

The stepwise linear regression models for the SMW, TUG, and STR are shown in Table 3. The FSE explained 45% or more of the variance in the SMW, TUG, and STR scores, suggesting that the FSE is the major determinant of these performance scores. Mechanical variables, such as the BMI and strength combined, contributed less than 15% to the models. Specifically, a model of the FSE, QUAD, BMI, and PSE explained 62.0% of the subjects' SMW scores. The last 3 variables added a total of 11.4% to this model. The significant factors explaining the variance in the TUG scores included the FSE, which contributed 51.7% of variance to the TUG scores, and the QUAD and BMI measures, which each added another 6% to the TUG score variance. Finally, a model of the FSE and HAM explained 52.7% of the STR scores, with the FSE contributing 45.7%.

With the 2 strength variables (QUAD and HAM muscle strength) excluded, an analysis of the multicollinearity of these regression models showed little correlation between the independent variables, with the smallest tolerance value at .75 and all others above .83, indicating that the regression models were not affected by interrelationships between explanatory variables.³⁶ To investigate whether QUAD and HAM could be used interchangeably because of the high correlation coefficient (r=.794), we ran each regression for the physical performance measures without one strength variable. In all cases, the subsequent regression model substituted the QUAD with the HAM or vice versa. Each resultant model explained the same amount of variance (within 5%) of the dependent variable as the original regression model presented here.

Analysis of potential outliers with the centered leverage values determined from residual statistics identified no extreme cases (values below 0.21) for the SMW and TUG regressions. For the STR regression, one extreme case was identified: a centered leverage value of 0.52, or 3.2 standard deviations from the mean, for a subject requiring 28 seconds to complete the test. Because this STR score is clinically possible,³⁷ we retained the data from this subject. Furthermore, analysis of the regression excluding this outlier did not affect the variables selected or the order of variables in the STR regression.

Finally, to confirm that mechanical variables did not share some variance with the psychosocial questionnaires in our regression models, we performed a stepwise linear regression for the SMW, TUG, and STR with only mechanical variables (BMI and strength) as independent variables. A model of HAM and BMI measures explained 30.2% of the SMW scores. Similarly, a model of HAM and BMI measures explained 36.7% of the TUG scores and 32.1% of the STR scores. In all cases, the BMI contributed less than 10%.

Discussion

The purpose of this study was to evaluate the relative contributions of psychosocial and mechanical factors to physical performance measures in people with knee OA. Self-efficacy explained much of the variance in performance in people with knee OA, with contributions from knee strength and body weight as well. The implications for physical therapist practice are that interventions in this population should aim to improve an individual's confidence in performing physical tasks in addition to mechanical strategies, such as weight loss and strengthening. These findings provide some evidence to support the clinical experience that psychosocial issues have a role in determining performance. Figure 2 depicts a proposed theoretical framework based on the results, showing that both mechanical variables and self-efficacy are related to physical performance in people with knee OA. Readers should be cautioned that the results of this cross-sectional study provide only an overview of the variables that influence performance. Longitudinal research is needed to determine whether variables such as self-efficacy can be manipulated to improve performance in people with knee OA.

Figure 2.

Theoretical framework identifying self-efficacy, strength, and body weight as factors related to physical performance in people with knee osteoarthritis (OA). Functional self-efficacy contributes at least 45% to the variance in performance of walking, stair climbing, and transferring in people with knee OA. Strength and body weight contribute 10% to 15% to the variance in performance.

Self-efficacy is the confidence that people have in their abilities to perform a specific task.²⁵ Bandura⁸ contended that how people behave is better predicted by their beliefs about their capabilities than by what they are actually capable of accomplishing. In addition to past successful performance, sources of self-efficacy include persuasion and observation of others. Persuasion refers to the act of verbally encouraging people that they possess the capabilities to master a given task.⁸ Observing another individual succeed raises the observer's beliefs that they, too, possess the capabilities to master comparable activities.⁸ Therefore, it is possible to determine self-efficacy for a task without considering past performance. In some cases, such as first-time child birth, past performance may not be available. From such a perspective, an individual's level of self-efficacy could be completely independent of his or her capabilities for performance.

Self-efficacy for physical tasks, measured by the FSE subscale of the Arthritis Self-Efficacy Scale, had the strongest relationship with all physical performance measures. The FSE also explained at least 45% of the variance in the SMW, TUG, and STR scores, suggesting that the more certain people were that they could complete physical tasks, the better they performed in walking, transferring, and stair climbing. Other studies also have highlighted the role of self-efficacy in people with knee OA. Gaines and colleagues⁷ found a significant relationship (r =−.559) between the FSE and self-reported performance in 29 women with knee OA. No significant relationship was found in the 14 men who participated, perhaps because of the small sample size. In the study by Gaines and colleagues,⁷ performance was self-reported as opposed to observed. Harrison³⁸ found that balance and self-efficacy explained 42% of the variance in physical performance, whereas self-efficacy and pain explained 74% of the variance in self-reported function. Sharma and colleagues⁹ showed that high levels of self-efficacy resulted in decreased odds of poor observed performance of a sit-to-stand task over the span of 3 years in 257 people with knee OA. It is important to note that the self-efficacy scores in these other studies^7,9,38 are much lower (indicating poorer levels of self-efficacy) than those obtained in our sample. Although these studies demonstrated that self-efficacy was important in performance, we specifically compared the relative contributions of psychosocial and mechanical variables to performance.

Although depression and anxiety are strong predictors of pain and function among people with knee OA,^5,39 the subjects in this study did not demonstrate anxiety or depression scores appreciably different from those in the general population. Our sample had a mean score of 63 (SD=16) on the STAI. In comparison, older adults with generalized anxiety score approximately 93 on the STAI, whereas older adults who are healthy score approximately 58.²⁹ Similarly, our sample had a mean score of 10 (SD=9) on the CES-D Scale; studies of subjects who were younger and healthy demonstrated scores of approximately 7,²⁶ and people with depressive symptoms over the preceding week had scores of over 16.²⁶ As a result, the subjects involved in this research study appear to be unique compared with those in other studies. We speculate that the high level of education in our sample may have contributed to healthier levels of depression and anxiety.⁴⁰ Although our correlations between the STAI or the CES-D Scale and the performance measures were weak (r =.12–.37), several studies^23,39,41 have highlighted a role for depression and anxiety in the perceived functional status of people with knee OA. It is possible that depression and anxiety mediate the relationship between perceived and performed physical abilities.

Our study sample included subjects who were, as a group, overweight. Previous studies of knee OA^30,42 included subjects with BMIs higher than those of our sample; our findings are limited to subjects with knee OA and relatively lower levels of obesity. In this study, the correlations between BMI and the SMW or the TUG were moderate, suggesting that the less obese a subject, the greater the distance that he or she can walk and the faster that he or she can transfer. In addition, BMI contributed 3.5% to the SMW score, suggesting that high body weight has some relationship to the performance of physical tasks, although perhaps not as much as anticipated. One other study showed similar results: Sharma and colleagues⁹ found that BMI did not relate to the performance of a sit-to-stand transfer task in 257 people with knee OA and an average BMI of 30.5.

The absolute values of strength and the ratio of QUAD to HAM muscle strength (0.52±0.19) matched the findings of other researchers.^13,31,32,43 Unfortunately, we found a paucity of established normative strength data, particularly for isokinetic knee flexion and extension in older men and women^44,45; thus, a comparison with normative data was not possible. Our results demonstrated that knee muscle strength had a good relationship with performance (r =.47–.52) in people with knee OA. In addition, QUAD or HAM strength contributed to the variance of how well our subjects performed physical tasks, suggesting that the stronger the subjects, the better their performance. Because of the strong relationship between QUAD strength and HAM strength, these variables could be used interchangeably in the regression models. Supporting our findings, a longitudinal study⁹ also demonstrated that knee strength contributed to the performance of physical tasks, but to a lesser degree than psychosocial factors, such as self-efficacy. Strength may be a better determinant of performance if the absolute values of strength are lower, perhaps at a level below a critical threshold for transferring, walking, and stair climbing. Further investigation would be necessary to elucidate whether strength is a more important determinant when absolute values of strength are low. In addition, most studies of strengthening for knee OA have included functionally relevant exercises, such as stair climbing and walking,^14,46–48 making it impossible to discern the contribution of strength versus that of self-efficacy to performance. That is, practicing walking may strengthen knee muscles and may provide a source of experience that can promote improved self-efficacy. These study designs make it impossible to elucidate whether improvements in performance are the result of improved strength or improved self-efficacy, or both. Nevertheless, knee strengthening needs to be an important component of treatment for people with knee OA and our results suggest that the ideal treatment for knee OA also should consider self-efficacy.

On the basis of our findings, therefore, physical therapists need to consider that treatment for knee OA would be improved by including strategies that increase self-efficacy in addition to strengthening knee muscles and producing weight loss. Strategies to improve self-efficacy for physical tasks in people with knee OA exist; for example, the Arthritis Self-Management Program involves 6 weekly, 2-hour sessions taught by a trained layperson and covering pathophysiology, exercise, relaxation, appropriate use of joints, medications, patient-physician communication, and problem solving.⁴⁹ People who had knee OA and participated in this program experienced a 15% to 20% decline in pain, a reduced number of physician visits, and improved self-efficacy, even over a 4-year follow-up period.^50,51 However, no study evaluated performance after this program, a direction for future study. It is theorized that education and practice mediate behavior changes in people.²³ These findings reinforce the pivotal role of physical therapists in providing practice and comprehensive education to promote self-management for people with knee OA. The role of self-efficacy in performance likely is substantial.

Some limitations of this study must be considered. In terms of generalizability, most characteristics of our study sample were typical of those of other study samples of subjects with knee OA. The gender distribution (59% female) and strength profiles were typical of people with knee OA. However, the mean score of 7 seconds (SD=3 seconds) on the TUG was consistent with literature indicating that community-dwelling women perform the test in between 6.0 and 11.2 seconds.⁵² It is possible that mechanical factors, such as strength and BMI, did not relate more strongly to performance on the TUG, SMW, and STR because our subjects performed well, that is, similar to community-dwelling subjects. Unlike the subjects in other studies, our subjects did not appear to have scores on the CES-D Scale or STAI that were appreciably higher than those in the general population. The subjects were overweight but to a lesser degree than in other studies. The Western Ontario and McMaster Universities Osteoarthritis Index scores (Tab. 1) indicated that our subjects had relatively lower levels of pain and impairment. The self-efficacy scores were considerably higher than those obtained in other studies of people with knee OA.^9,38 Thus, our sample likely was composed of subjects with mild to moderate knee OA. Different results may be found in samples with more severe knee OA.

No information about the duration of illness was recorded; thus, the subjects may have had acute or chronic disease. However, the recruitment process of obtaining a diagnosis was lengthy (mean of 3 months between initial contact and participation), suggesting that the subjects had chronic disease. The same tester recorded scores for all measurements. This tester was aware of the dependent and independent variables because all of the independent variables were considered important at the outset of the study. The order of measurements was not randomized, a factor that may have introduced bias into this study. However, the order of measurements was necessary to ensure adequate rest periods between physical tasks and to ensure that strength measures did not create fatigue before other performance measures. The reliability of data obtained with the measures and questionnaires was not established in this study because the reliability and validity of data obtained with the measures and questionnaires in studies of people with arthritis have been reported elsewhere. Finally, investigators are encouraged to include at least 5 subjects, ideally 10 to 15 subjects, for every independent variable included in regression analyses.⁵³ In this study, 8 independent variables were used (5 psychosocial and 3 mechanical), suggesting that at least 40 subjects were required, but 80 to 120 subjects would have been ideal. Our study included 54 subjects, a factor that may have resulted in an enhancement of type I errors.

Conclusion

Physical therapists, we believe, should use outcome measures to document a change in status in the mobility of people with knee OA. However, understanding what the scores mean also can guide clinical treatment. Self-efficacy, or a person's confidence in his or her ability to complete a physical task, explained much of the variance in data obtained with the 3 physical performance measures in our group of subjects with mild to moderate knee OA. Body weight and strength also explained some of this variance. Therefore, strategies to improve physical performance in people with knee OA should include not only a combination of mechanical treatments for weight loss and strengthening but also psychosocial interventions aimed at improving self-efficacy. In addition, physical therapists evaluating the significance of the SMW, TUG, and STR scores in subjects with knee OA should note that a large part of each score reflects an individual's self-efficacy for physical tasks. Further research is necessary to develop concrete strategies for physical therapists to use in aiming to increase the level of confidence of their subjects with knee OA.