Self-Efficacy and the Progression of Functional Limitations and Self-Reported Disability in Older Adults With Knee Pain

Abstract

Objectives. The prospective relationships between self-efficacy beliefs, in conjunction with measures of knee pain and knee strength, and subsequent decline in both physical performance and self-reported disability among older adults with knee pain were examined.

Methods. In this prospective epidemiological trial, 480 men and women aged 65 years and older who had knee pain on most days of the week and difficulty with daily activity were followed for 30 months.

Results. There was a significant interaction of baseline self-efficacy with baseline knee strength in predicting both self-reported disability and stair climb performance. Participants who had low self-efficacy and low strength at baseline had the largest 30-month decline in these outcomes.

Discussion. These data underscore the important role that self-efficacy beliefs play in understanding functional decline with chronic disease and aging. Self-efficacy beliefs appear to be most important to functional decline in older adults when they are challenged by muscular weakness in the lower extremities.

WITH the rapidly growing number of older adults in North America, there is widespread agreement that functional limitations and disability are among the most significant challenges to confront health care in the new millennium (Fried and Guralnik 1997). Seeman, Unger, McAvay, and Mendes de Leon 1999 have proposed that deciphering the etiology of functional disability is key to maximizing the compression of morbidity and disability to the latter stages of life. Although the prevalence of functional limitations and disability has been linked to pathology and physical impairments (Verbrugge and Jette 1994), there is growing support for the position that decline in these parameters is also related to older adults' self-efficacy beliefs.

Tinetti, Mendes de Leon, Doucette, and Baker 1994, employing a Falls-Efficacy Scale, provided some of the first evidence that control beliefs such as self-efficacy are important in understanding activity restrictions in older adults. This relationship held even after Tinetti and colleagues controlled for the history of recent falls and injury. In 1996, the Falls-Efficacy Scale was found to be prospectively related to decline in self-care tasks; however, the effect was strongest for those who experienced the greatest decline in performance-related disability (Mendes de Leon, Seeman, Baker, Richardson, and Tinetti 1996). In that same year, Rejeski and his colleagues in a study of older adults with knee osteoarthritis (OA) found that self-efficacy beliefs specific to the performance of functional tasks were predictive of performance on these tasks even after controlling for markers of pathology and fitness (Rejeski, Craven, Ettinger, McFarlane, and Shumaker 1996). In more recent studies, self-efficacy has been linked both cross-sectionally (Foldvari et al. 2000) and prospectively (Seeman et al. 1999) to declines in self-reported disability. However, in contrast to the work by Rejeski and his colleagues (1996), Seeman and her colleagues (1999) did not find a link between self-efficacy and performance-based functional assessments. This latter discrepancy may well be related to differences in measurement. That is, Seeman and her colleagues employed a broad-based measure of efficacy (ability to arrange transportation, living arrangements, safety, and one's own productivity), whereas Rejeski and his colleagues assessed self-efficacy specific to the tasks being performed. Bandura 1997 has noted that any mismatch between the assessment of self-efficacy and the behavior of interest reduces the strength of the relationship between the two.

Our purpose in the present study was to explore the 30-month prospective relationships between self-efficacy beliefs for functional tasks and declines in performance-related function as well as self-reported disability among older adults with knee pain. As conservative estimates for the effect of self-efficacy on these outcomes, model testing included measures of knee pain and knee strength.

Methods

Overview of Study Design

The Observational Arthritis Study in Seniors (OASIS) was a 30-month prospective investigation on the progression of physical disability in older adults with knee pain. Researchers recruited participants using telephone-based interviews of individuals drawn from a commercial list of people aged 65 or older residing in Winston-Salem, NC, and surrounding counties. Names on the commercial list were randomly selected and received a mailed brochure followed by a telephone call. During phone calls, a brief screening questionnaire was administered and persons who met our initial criteria were asked to provide informed written consent. The original population consisted of 17,790 older individuals. We contacted 11,191 in this pool and identified 940 potentially eligible persons. Of this group, 208 expressed no interest in the study. Of the 732 remaining individuals, 480 passed the detailed screening for inclusion and were enrolled in the study, for a response rate of 66%. Self-report and measurement data were collected at baseline and at 15- and 30-month visits. We asked participants who refused to return for follow-up visits to complete questionnaires in their residence and return the completed questionnaires via mail.

Participants

A total of 480 participants met eligibility criteria and completed baseline evaluations. The eligibility criteria for participation were (a) aged 65 years or older, (b) knee pain on most days, and (c) difficulty with at least one of the following because of knee pain: walking a quarter mile; climbing stairs; getting in and out of a car; rising from a chair; lifting and carrying groceries; getting out of bed; getting out of the bathtub; or performing shopping, cleaning, or self-care activities. Potential participants were excluded if they (a) were moving from the area within 3 years, (b) were under hospice care, (c) were receiving active treatment for cancer (other than skin cancer), (d) had shortness of breath or chest pain at rest, (e) had a score of less than 24 on the Mini-Mental State Examination (Folstein, Folstein, and McHugh 1975), (f) had rheumatoid or psoriatic arthritis, or (g) were currently participating in another study.

Measures

Stair climb task.

The stair climb task involved ascending and descending a set of five stairs that was equipped with a handrail and a platform at the top. Participants were asked to climb to the top of the steps with their left hand on the rail as quickly as they could. Once both feet were on the top platform, they were instructed to turn around quickly counterclockwise, to regrasp the handrail, and to climb down as quickly as possible. The task was scored as the total time to go up and down the stairs. As described by Rejeski, Ettinger, Shumaker, James, and colleagues (1995), the task has excellent 2-week test–retest reliability (.93) and correlates in expected directions with knee strength (−.58), VO_2peak (−.37), and self-reported ambulatory disability (.38). Performance on the stair climb has also been found to be sensitive to change with physical activity interventions (Ettinger et al. 1997). This measure of physical function was chosen because it can be easily replicated in a clinic environment.

Self-reported disability.

This questionnaire consisted of items that assess levels of difficulty with activities of daily living (Rejeski, Ettinger, Shumaker, James, et al., 1995). Participants were asked, "How much difficulty, if any, did you have with each of these activities? Think about the past month." For each activity, responses were measured on a 5-point Likert scale ranging from 1 (no difficulty) to 5 (unable to do). We obtained a composite disability score by averaging the scores on 10 items covering the following activities: doing light housework, preparing meals, participating in community activities, managing money, visiting with relatives or friends, using the telephone, dressing, taking care of a family member, eating, and shopping. Principal component analysis revealed that these items could be treated as a composite score (all loadings were in excess of .50). At the baseline exam, these items had a Cronbach alpha coefficient for internal consistency of .83. Short-term reliability studies (1–2 weeks) from our center have produced correlations of .80 or higher.

Sociodemographic variables.

Variables representing gender and age (measured in years) were constructed from self-reports.

Radiographic evidence of knee OA.

Radiographic evidence of knee OA was defined according to criteria described by Felson and colleagues 1997, where an osteophyte score of 2 or greater was shown to correlate best with clinical evidence of knee OA. Specifically, a dichotomous variable was generated consisting of either no radiographic evidence of OA in both knees, (osteophyte scores of less than 2) or radiographic evidence of OA in either knee (osteophyte score of 2 or greater).

Knee pain.

We assessed knee pain using the ambulatory and transfer pain intensity subscales from the Knee Pain Scale (KPS; Rejeski, Ettinger, Shumaker, Heuser, et al. 1995). Each subscale has three items that are rated on a 6-point Likert-type scale with the following anchors at each integer: no pain (1), mild pain (2), discomforting pain (3), distressing pain (4), horrible pain (5), and excruciating pain (6). The ambulatory items include climbing up and climbing down a flight of stairs and walking a short distance (1 block), whereas the transfer items involve getting into or out of a bed, chair, and car. Means are calculated for the three items on each subscale and can range from 1 to 6 with higher scores indicating more severe levels of pain. The KPS has good convergent validity, and confirmatory factor analyses have supported the creation of separate scores for ambulation and transfer (Rejeski, Ettinger, Shumaker, Heuser, et al. 1995). Cronbach alpha reliabilities and test–retest reliabilities for the subscales are all in excess of .80 (Rejeski, Ettinger, Shumaker, Heuser, et al. 1995).

Knee strength.

We used a Kin-Com 125E isokinetic dynamometer to assess concentric knee extension strength. The reliability and validity of this device has been well documented (Farrell and Richards 1986; Harding, Black, Bruulsema, Maxwell, and Stratford 1988). Gravity effect torque was calculated based on the participant's leg weight at a 45° angle. An angular velocity of 30°/s was used for all tests. The activation force for each muscle group was set at 50% of maximal voluntary isometric contraction and was based on reports that the amount of activation force significantly influences the magnitude of the average force recorded (R. C. Jensen, Warren, Laursen, and Morrissey 1991). Knee extensors were tested through a joint arc from 90° to 30° (0° = full extension). We deleted the first and last 10° to account for the acceleration and deceleration of the dynamometer at the ends of the range of motion and also to account for possible inconsistent effort (Kramer, Vaz, and Hakansson 1991). Two maximal reproducible trials were averaged, and the maximum number of trials did not exceed six (Kramer et al. 1991). The most affected leg (i.e., the knee that produced greater pain) was used for each test. Average force (N) exerted was divided by body mass (kg) to provide a measure of relative strength (N/kg).

Self-efficacy.

The self-efficacy measure employed in this investigation was collected in conjunction with the stair climb task (Rejeski, Ettinger, Shumaker, James, et al., 1995). After performing a practice trial, participants were presented with a confidence ladder drawn on a large poster board. This ladder had 10 steps ranging from 0 (completely uncertain) to 10 (completely certain). Participants were asked to rate the level of certainty that they could complete the stair climb task 2 times, 4 times, 6 times, 8 times, and 10 times without stopping. We calculated baseline self-efficacy scores by summing confidence ratings across the five levels of effort and multiplying this result by 2 to produce a score that ranged from 0 to 100. This hierarchical measurement protocol is consistent with the standard protocol developed by Bandura 1986 to measure task-specific self-efficacy beliefs. This particular scale has shown to have construct validity in two previous publications (Rejeski et al. 1996; Rejeski, Ettinger, Martin, and Morgan 1998).

Statistical Analyses

We used mixed effects, repeated-measures analysis of covariance to relate each outcome variable (stair climb time and self-report disability) to baseline measures of knee strength, knee pain, and self-efficacy (Laird and Ware 1982). We used pain intensity during ambulation to predict stair climb performance, and we used pain intensity during both ambulation and transfer to predict self-report disability. We also explored potential interactions between knee strength, knee pain, and self-efficacy and included significant terms in the final models. Covariates we used to adjust mean levels of stair climbing performance and self-reported disability included gender, age, knee pain intensity, an indicator of radiographic knee OA, concentric knee extension strength, and self-efficacy.

These models employed a popular parameterization used for longitudinal, observational studies (Diggle, Liang, and Zeger 1994) and permitted the estimation of cross-sectional versus longitudinal effects for each predictor variable. In this article, we focus on how several variables collected at baseline alter the effect of aging (i.e., follow-up time) on change in each outcome.

Results

Follow-up data for 16 OASIS participants were excluded from the analyses because the participants had knee replacement surgery at some point during follow-up. Furthermore, 21.6% of the remaining participants were too ill, had knee surgery, or did not complete questionnaires at the final assessment, whereas an additional 8.6% were either unable or unwilling to perform the stair climb performance task at this point in time. Thus, at the 30-month visit, 364 participants completed the self-report disability questionnaire, whereas 324 completed the stair climb task.

Descriptive Statistics

As shown in Table 1 , the average age of the OASIS participants was 71.82 years. There was nearly equal representation by men and women, and 82.5% were Caucasian. The most common comorbidities were hypertension and obesity, with 41.3% having a body mass index equal to or more than 30. At entry into the study, participants expressed a level of knee pain that can be described as mild (a rating of 2) to uncomfortable (a rating of 3), and they had noticeable deficits in physical function. Specifically, their concentric knee strength was approximately 25% lower that what one might expect in this age group (Messier, Loeser, Hoover, Semble, and Wise 1992). There was substantial variability (SD = 4.70 s) at baseline in time to complete the stair climb task, and the mean of 10.04 s was comparable to what we had observed in other community trials of older adults with knee osteoarthritis (Ettinger et al. 1997). Across the 30 months of the study, the mean decline in self-reported disability was 0.13 and in stair climb time was 1.99 s.

Relationships Between Predictors and Outcomes at Baseline and Longitudinal Analyses

At the baseline assessment, all of the predictor variables had significant (p < .001) bivariate relationships in the expected direction with both self-reported disability and performance on the stair climb task. Self-reported disability correlated .37 with ambulatory knee pain, .37 with transfer knee pain, −.23 with knee strength, and −.33 with stair climb self-efficacy. Performance on the stair climb correlated .32 with ambulatory knee pain, .25 with transfer knee pain, −.63 with knee strength, and −.53 with stair climb self-efficacy. The strongest relationships were found for knee strength and self-efficacy with stair climb time. As expected, better performance was observed in participants who had stronger knees or higher self-efficacy.

Longitudinal models were developed for self-reported disability and stair climb performance with the mean levels of each outcome at each visit adjusted for age, knee pain intensity, x-ray scores, concentric knee strength, and self-efficacy. On the basis of the results from repeated-measures analyses, the equation for 30-month change in self-reported disability = (−0.5666) + (−0.0894 · baseline ambulatory pain intensity) + (0.0333 · baseline transfer pain intensity) + (0.1477 · knee strength) + (0.0066 · self-efficacy) + (−0.0017 · knee strength/self-efficacy interaction). In this model, the only significant predictors were strength (p < .05), self-efficacy (p < .01), and the interaction term for strength by self-efficacy (p < .05). The equation for 30-month change in stair climb performance = (−14.5636) + (−0.4374 · baseline ambulatory pain intensity) + (3.3152 · knee strength) + (0.1423 · self-efficacy) + (−0.3328 · knee strength/self-efficacy interaction). As with self-reported disability, the only significant predictors were strength (p < .001), self-efficacy (p < .001), and the interaction term for strength by self-efficacy (p < .001). Because the knee strength by self-efficacy interactions qualified interpretations of main effects, Table 2 provides estimates of least-square means characterizing the interaction effect from the repeated-measures analysis of covariance. The means represent estimated scores for baseline strength and self-efficacy at the 25th and 75th percentiles for each variable. The interpretation was consistent for both outcomes; that is, self-efficacy was related to greater decline in physical function when knee strength was low.

Discussion

The results of this study reinforce the position that self-efficacy beliefs are important in understanding functional decline that occurs with chronic disease and aging (Mendes de Leon et al. 1996; Rejeski et al. 1996; Seeman et al. 1999; Tinetti et al. 1994). To our knowledge, the current report provides the first prospective data available on older individuals who have compromised function due to knee pain and is unique in that the interaction effect for self-efficacy and strength holds for both self-report and performance-based outcomes. Although a previous prospective investigation by Seeman and her colleagues (1999) did include performance measures, self-efficacy was predictive of self-reported ADL disability only. One reason for the discrepancy between these latter data and those reported in the current study may well be that the self-efficacy measure employed by Seeman and her colleagues (1999) was not targeted to performance-based tasks; that is, items on the scale tapped confidence associated with transportation, living arrangements, safety, and productivity. Correspondence in measurement is an important issue when researchers are attempting to understand the predictive and explanatory power of self-efficacy on outcomes such as functional limitations and disability (cf. Courneya 1994).

It is important to underscore the fact that, in the present study, a significant interaction between strength and self-efficacy qualifies interpretation of the main effect for self-efficacy. Specifically, baseline levels of self-efficacy are important in predicting decline in physical function, but only for participants who had poor baseline knee strength. This finding is consistent with existing data on self-reported disability showing that self-efficacy is predictive of decline in function across time when objective function is compromised at the time of baseline assessment (Mendes de Leon et al. 1996). Collectively, these results support the postulate that self-efficacy is most protective when people are challenged by deteriorating function (Bandura 1986). Furthermore, the findings from the present study corroborate the important role of control beliefs in understanding both functional limitations and self-reported disability that are exacerbated by clinical pain (M. P. Jensen, Turner, Romano, and Karoly 1991).

Some readers may have questions concerning the external validity of the study results. We want to emphasize the fact that we invested considerable resources into identifying a population of older adults in the community who had some disability due to knee pain. We did not include individuals whose performance on the assessment protocol would have been compromised from the very beginning. Of our target population, we were able to recruit 66% into the study. Moreover, we employed statistical models that allowed us to account for the effects of missing data. Hence, we express confidence in the external validity of this study, while at the same time underscoring the characteristics of the target population; that is, older adults in the community who have knee pain; express mild to moderate levels of disability; and are not compromised by low cognitive function, extreme physical disability, or severe chronic disease. We encourage additional investigations on this important topic with older adults who have severely compromised health status for one reason or another, yet we caution investigators to use performance measures designed for these subgroups that will help circumvent problems that arise from floor effects.

In summary, the results of this study underscore the clinical importance of self-efficacy beliefs to the progression of functional limitations and self-reported disability in older adults who are classified as high risk because of compromised function. The reader should keep in mind that the participants in this study had knee pain; thus, the findings may not generalize to individuals with disabilities resulting from other medical problems. Programs designed to impede the decline of physical function in this population should be multifaceted. For example, we would hypothesize that the demonstrated efficacy of exercise programs on physical function in older adults with knee OA (Ettinger et al. 1997) could be enhanced by coupling physical rehabilitation with programs of pain management and self-efficacy enhancement (Rejeski et al. 1998).

Table 1.

Descriptive Statistics on Older Adults With Knee Pain and Baseline Means and Standard Deviations for Predictor and Outcome Variables

Variable	n	M or %	SD	Range
Gender (% female)	480	51.0%
Age (years)	480	71.82	5.00	65–88
Race (% caucasian)	480	82.5%
Education (% ≥12 years)	480	58.0%
Live alone (% yes)	476	27.9%
Hypertension (% yes)	474	38.8%
Cardiovascular disease (no. of symptoms)	479	0.64	0.84	0–4
Diabetes (% yes)	473	14.4%
COPD (% yes)	461	12.6%
BMI (kg/m2; obesity = BMI≥30)	480	29.6	5.22	18.0–62.2
Osteophyte score ≥2 (% yes)	463	51.6%
Ambulatory pain intensity score	471	2.48	0.70	1–6
Transfer pain intensity score	473	2.38	0.67	1–6
Knee strength (N/kg)	472	2.34	1.14	0.32–7.12
Self-efficacy for stair climb	476	72.65	28.55	0–100
Stair climb time (s)	473	10.04	4.70	4.1–40.1
Self-reported disability with ADLs/IADLs	464	1.33	0.39	1.0–3.7

Notes: ADL = activities of daily living; BMI = body mass index; COPD = chronic obstructive pulmonary disease; IADL = instrumental activities of daily living.

Table 2.

Least Squares Mean Estimates Characterizing the Strength by Self-Efficacy Interaction Term for Self-Reported (SR) Disability and Stair Climb Performance Obtained From Repeated-Measures Analysis of Covariance

	Baseline Percentile			M
Outcome	Self-Efficacya	Strengthb	Baseline	30 Months	Change
SR disability	25	25	1.45	1.65	0.20
	25	75	1.39	1.49	0.10
	75	25	1.26	1.34	0.08
	75	75	1.20	1.31	0.11
Stair climb (s)	25	25	11.80	15.95	4.15
	25	75	9.56	11.03	1.47
	75	25	10.21	11.51	1.30
	75	75	7.98	9.20	1.22

Notes: The sample consisted of 480 older adults with knee pain. Estimates were obtained at the mean levels of the other predictors in the model including baseline age, pain intensity, x-ray scores, and gender. For categorical predictors, these estimates were equivalent to taking an average based on the proportion of participants with the characteristic in the initial baseline sample.

a

The 25th percentile for self-efficacy = 50 and 75th percentile = 100.

b

The 25th percentile for strength = 2.6 N/kg and 75th percentile = 4.2 N/kg.

Support for the Observational Arthritis Study in Seniors (OASIS) was provided by the National Institute on Aging (R01-AR42388) and the Geriatric Clinical Research Center (GCRC; M01-RR00211). Preparation of this manuscript was supported by the Claude D. Pepper Older Americans Independence Center of Wake Forest University (5P60AG10484). The authors acknowledge the contributions made by Walter H. Ettinger, Jr., in the development and operations of the OASIS study.