Advertisement

Abstract

Background and Purpose. Physical activity and exercise play a critical role in the management of arthritis. Understanding the factors affecting physical activity and exercise behavior is a necessary first step toward identifying the needs of, and intervention strategies for, people with arthritis. The purpose of this study was to identify factors affecting physical activity and exercise behavior in urban subjects with osteoarthritis (OA) and rheumatoid arthritis (RA). Subjects. Seventy-two consecutive subjects were recruited from the rheumatology clinic at a large urban public hospital. The sample was predominantly African American (92%), female (87%), and not working (90%). The subjects’ average age was 60.9 years (SD=13.9, range=30–90). Methods. Time per day spent sitting or lying down and time per week spent in exercise, leisure, and household activities were determined by individual interview. Self-efficacy, outcome expectations, disability, pain, body mass index, and social support were measured as possible explanatory factors. Results. The average daily total activity time was 3.1 hours. Household and leisure activities accounted for 85% of that time. Explanatory factors for physical activity behavior were not the same for subjects with OA and RA, despite similar between-group characteristics. Self-efficacy was present in all of the significant explanatory models. Discussion and Conclusion. The results indicate that factors that affect physical activity behavior among urban and predominantly African-American adults are dependent upon the type of physical activity and are different for people with OA and RA. Self-efficacy was the most consistent explanatory factor. [Greene BL, Haldeman GF, Kaminski A, et al. Factors affecting physical activity behavior in urban adults with arthritis who are predominantly African-American and female.

Arthritis is the leading cause of physical disability in the United States and affects nearly 43 million people.1 The most common form of arthritis is osteoarthritis (OA), and a second common form is rheumatoid arthritis (RA).2 Physical therapists frequently prescribe exercise and physical activity for the promotion of health and wellness, the prevention of secondary conditions, and the rehabilitation of people with arthritis.3,4 The critical role of exercise and physical activity is evidenced by recommendations of the American College of Rheumatology for muscle strengthening and aerobic conditioning exercises for people with hip and knee OA5 and the Centers for Disease Control and Prevention recommendation that every US adult, including those with disabilities, get at least 30 minutes of moderate-intensity physical activity on most days of the week.6

Despite the known benefits of physical activity and exercise, physical inactivity is a national problem, and people with arthritis are more likely to be physically inactive and either overweight or obese than are people without arthritis.7 On the basis of data from the Behavioral Risk Factor Surveillance System, only 24% of people with arthritis report engaging in physical activity at the recommended level.8 The proportion engaged in physical activity at the recommended level is lower and the reported activity restrictions are higher in women of African-American or Hispanic descent.9,10

A number of theoretical models have been studied and used as a basis for understanding physical activity and exercise behavior.11,12 The Social Cognitive Theory (SCT) describes a dynamic and reciprocal triad in which behavior, personal factors, and environmental factors all interact.13 According to the SCT, an individual’s behavior, such as participation in physical activity, is influenced by physical or social environmental factors and personal factors. Systematic reviews of the literature on predictors of physical activity reveal that the SCT constructs of self-efficacy, outcome expectations, and social support, among others, are important predictors of physical activity behavior for the general population.11,12 Self-efficacy refers to the confidence that an individual has in performing specific behaviors; higher self-efficacy in a person’s ability to participate in physical activity and exercise is predictive of higher levels of physical activity.1113 A stronger belief in the benefits of physical activity and greater social support to participate in physical activity also are predictive of higher levels of physical activity.11,12

Few research studies have examined predictors of physical activity behavior for people with arthritis, and none of these studies was done with a predominantly minority population. Although physical activity encompasses a variety of different types of activities, few studies have investigated more than 1 or 2 types of physical activity. A cross-sectional study of 100 subjects with either OA or RAshowed that the strongest direct predictor of participation in exercise was the perceived benefit of exercise.14 A recent prospective study of subjects with RA showed that predictors of exercise behavior at 6 months after exercise onset were past history of exercise and the rheumatologists’ current exercise behavior.15 Secondary analysis of data collected from 120 subjects with OA and RA in a randomized controlled trial of aerobic exercise demonstrated that the maintenance of exercise behavior was associated with measures of physical fitness, depression, anxiety, and prior exercise behavior.16 The need for a greater understanding of factors associated with physical activity in people with arthritis has been recognized.8,17 Understanding the factors that affect physical activity behavior is a necessary first step toward identifying the needs of, and potential intervention strategies for, this population. The purpose of this study was to answer the following questions: (1) What factors explain physical activity and exercise behavior in urban adults with OA and RA who are predominantly African-American and female? and (2) Are the factors that explain physical activity and exercise behavior the same for people with OA and RA in this population?

Method

Subjects

Subjects were recruited from the rheumatology clinic at a large urban public hospital providing health care services to uninsured and underinsured populations. The hospital and clinic patient base is predominantly urban and African American. We attended the rheumatology clinic from May through June 2003. All of the charts of subjects attending the clinic each day were reviewed to identify potential subjects meeting the following inclusion criteria: older than 30 years of age, classified as having OA or RA according to American College of Rheumatology criteria,2 reported less than 1 hour of morning stiffness, and comprehended English. Ninety-eight consecutive subjects met the inclusion criteria and were asked to participate in the study. Nineteen subjects (19%) declined participation, and 79 subjects (81%) gave signed informed consent. Three subjects (4%) were used to pilot test the measures, bringing the final sample size to 76 subjects.

Four of the 76 participants were excluded from data analysis due to incomplete data. Of the 72 subjects who had complete data, 72% (52) had a primary diagnosis of RA and 28% (20) had a primary diagnosis of OA. Subjects who had both OA and RA (29%) were included in the RA group, as OA was assumed to be secondary to RA. All of the subjects had been diagnosed with arthritis for more than 1 year. The sample was predominantly African American (92%), female (87%), and not working (90%). The average age of the subjects was 60.9 years (SD=13.9, range=30–90). Nearly half of the subjects (43%) had not completed high school, 40% had completed high school, and 17% had attended or completed college. On average, subjects had 2 comorbid conditions. The most frequently reported comorbidities were hypertension (68.5%), diabetes (30.1%), depression (24.7%), and coronary artery disease (12.3%). Twenty percent of the participants were normal weight, 25% were overweight, and 55% were obese.

Data Collection

A cross-sectional research design was used. All data were collected in a single session. The subjects completed a demographic questionnaire and then underwent height and weight measurements. The order for measurement of the remaining variables (physical activity, self-efficacy, outcome expectations, disability, pain, and social support) was randomized by the roll of a die.

Outcome Variables

Physical activity was defined as “any bodily movement produced by skeletal muscles that results in energy expenditure.”18 Each subject was interviewed with the Physical Activity and Disability Survey (PADS).19 The survey was administered as a semistructured interview as described by the originators.19 First, the subject was asked, “Do you exercise?” If the response was “yes,” then the subject was asked to describe the exercise activity and to estimate the amount of time spent exercising during the preceding week. To assess leisure physical activities, the subject was asked, “Do you currently participate in any sports or recreational or leisure activities?” If the response was “yes,” then the subject was asked to describe the type of activity and to estimate the amount of time spent in the activity during the preceding week. To assess household activities, the subject was asked, “Are most of your indoor household activities done by you or by someone else?” If the response was that the subject did most of the activity, then the subject was asked to describe the activity and to estimate the amount of time spent in the activity during the preceding week. The subject was asked a similar question about outdoor household activities. A total weekly physical activity score was obtained by summing the exercise, leisure, and household activities in minutes per week and then dividing by 7 for a daily estimate. An intraclass correlation coefficient (ICC[3,3]) of .99 was obtained for interrater reliability of PADS scores throughout the study. An ICC of .85 has been reported in the literature for test-retest reliability of PADS Total Activity Scale scores.19 The concurrent validity of PADS scores has been supported by significant correlations of the subscale and Total Activity Scale scores with measures of cardiovascular fitness in people with chronic health conditions and physical disabilities.19 The predictive validity of the exercise subscale and Total Activity Scale scores has been supported by a significant group × time interaction following an exercise intervention in the same cohort.19

Physical activity also was measured as a subject’s report of the average amount of time spent sitting or lying down during waking hours per day and was assessed with a single question, “On average, how many hours per day are you sitting or lying down, not counting when you sleep at night?” This question was obtained from the PADS; however, to our knowledge, no reliability or validity information is available for this item. The score on this single question is separate from the PADS subscale and Total Activity Scale scores.

Explanatory Variables

Selection of the predictor variables was based on previous literature and was consistent with the SCT. Other than demographic variables, only modifiable variables were of interest. The predictor variables were self-efficacy, outcome expectations, disability, pain, body mass index (BMI), and social support.

Self-efficacy was measured with the Arthritis Self-Efficacy Scale (ASES) “other symptoms” subscale, consisting of a 6-item questionnaire.20 Responses were recorded on a scale from 1 to 10, with 1 meaning “very uncertain” and 10 meaning “very certain.” The total ASES score was determined by calculating the mean of the 6 ASES items; the higher the score, the greater the self-efficacy. An ICC(3,3) of 1.0 was obtained for interrater reliability of ASES scores throughout the study. The internal consistency of the ASES in this study was calculated with a standardized Cronbach alpha coefficient of .95. The test-retest reliability of the ASES “other symptoms” subscale was previously reported in the literature to be r=.90.20 The construct validity of the subscale scores was demonstrated by significant correlations with health status measures.20

Outcome expectations were defined as a subject’s beliefs about the effects of exercise and physical activity. Outcome expectations were measured with the Outcome Expectations for Exercise (OEE) Scale.21 The OEE Scale is a 9-item measure focusing on the subject’s perceived consequences of exercise. Subjects completed the OEE Scale themselves in a written format rather than in an interview as originally described by Resnick et al21 because, during pilot testing, the research team perceived that the subjects’ report of the benefits of exercise were biased when physical therapists, known to promote exercise, read a statement about exercise. The responses to each statement on the benefits of exercise ranged from 1 (“strongly agree”) to 5 (“strongly disagree”). The scores were averaged to determine a final score for outcome expectations; the lower the score, the greater the perceived benefits of exercise and physical activity. The internal consistency of the OEE Scale in this study was calculated with a standardized Cronbach alpha coefficient of .93. An ICC(3,1) of .82 was obtained for test-retest reliability between the interview format and the written format for the OEE Scale in this study. In a previous study,21 using structural equation modeling to assess the reliability and validity of OEE Scale scores in older community-dwelling adults, each of the items demonstrated significant correlations with the OEE Scale score (path coefficient=.69–.87), and 83% of the variance in the OEE Scale score was explained by the 9 items.

Disability was defined as the inability to perform personal and instrumental activities of daily living. Disability was measured with the Health Assessment Questionnaire (HAQ).22 The HAQ is a 20-question survey with a 4-point scale addressing the following activities of daily living performed within the preceding week: dressing and grooming, arising, eating, walking, performing hygiene, reaching, gripping, and performing outdoor activities. The response choices ranged from 0 (“without difficulty”) to 3 (“unable to do”). The highest score in each category was added to determine a total score, which was divided by 8 to determine a functional disability index. The final functional disability index score ranged from 0 (“minimal difficulty”) to 3 (“total disability”). An ICC(3,3) of .99 was obtained for interrater reliability of HAQ scores throughout the study. The internal consistency of the HAQ in this study was calculated with a standardized Cronbach alpha coefficient of .90. High test-retest reliability (r=.98) and construct, criterion, and predictive validity for subjects with arthritis were reported previously.2224

Pain was defined as an unpleasant physical sensation. Pain was measured with the visual analog scale (VAS) of the HAQ.22 A 15-cm line with the end points labeled “no pain” and “very severe pain” was used. The subjects marked on the line the highest level of pain experienced from the arthritis in the preceding week. We measured the distance (in centimeters) between the “no pain” end point and the line marked by the subject and then multiplied that value by .667 to convert the data to a 10-cm scale. An ICC(3,3) of 1.0 was obtained for interrater reliability of VAS scores throughout the study. Test-retest reliability values for both literate (r=.93) and illiterate (r=.71) subjects with arthritis were reported previously.25 The criterion validity of the VAS scores compared with scores on a verbal descriptor scale was reported to have correlations of .70 to .75 for subjects with arthritis.26

Body mass index is the relationship of weight (in kilograms) to height (in meters squared) and was used to determine whether a subject was underweight, normal weight, overweight, or obese. The World Health Organization defines weight status according to the following BMI categories: less than 18.5 kg/m2 as underweight, 18.5 to 24.9 kg/m2 as normal weight, 25.0 to 29.9 kg/m2 as overweight, and greater than 30.0 kg/m2 as obese.27 Subjects were weighed on a scale that we calibrated to a known weight at the start of each day of data collection. We converted pounds to kilograms prior to data analysis. The height of each subject was measured by 1 of the investigators using a tape measure mounted against a wall during the interview process. The tape measure was calibrated by comparing it to another tape measure. An ICC(3,3) of 1.0 was obtained for interrater reliability of BMI measurements throughout the study.

Social support was defined as the availability of someone to provide help or emotional support. Social support was measured with the Medical Outcomes Study (MOS) Social Support Survey, consisting of 19 items.28 The responses ranged from 1 (“none of the time”) to 5 (“all of the time”). The scores were summed, rescaled on a 100-point scale, and then averaged to determine a total score for social support; the higher the score, the greater the perceived social support. An ICC(3,3) of .99 was obtained for interrater reliability of MOS Social Support Survey scores throughout the study. The internal consistency of the MOS Social Support Survey in this study was calculated with a standardized Cronbach alpha coefficient of .95. Evidence of validity of the MOS Social Support Scale scores was reported previously by significant correlations with measures of loneliness (r=−.67), family functioning (r=.53), marital functioning (r=.56), and mental health (r=.45).28

Data Analysis

Data were analyzed with SPSS version 11.0.* Incomplete data were managed as follows: if more than 10% of the total data were missing, then the data for that subject were eliminated from the study; if more than 10% of the questions on a questionnaire were incomplete, then the data from that questionnaire were eliminated from analysis. As a result, data for 4 of the 76 subjects were excluded from the final data analysis, and the sample sizes among the measurers varied slightly. Results of the t test and chi-square analyses revealed no significant differences in values for demographic, predictor, or outcome variables between the data for the 4 excluded subjects and the data for the rest of the sample.

Descriptive statistics were calculated for all variables. Either a chi-square test or a t test was used to determine whether a difference in values for demographic, predictor, or outcome variables existed between subjects with a primary diagnosis of OA and those with a primary diagnosis of RA. The study-wide alpha level was kept at .05 through the use of the Bonferroni technique, resulting in per-comparison alpha levels of .004 for the t test and .02 for the chi-square test.29 Bivariate regression and correlation analyses were done to assess the relationship between each explanatory variable and each outcome. Forward stepwise multiple linear regression was used to build a model of explanatory variables that best explained the variance in physical activity. Variables with a correlation of .20 or greater with outcomes were placed in the linear regression model. Multicolinearity among the explanatory variables was assessed by calculating the tolerance level and bivariate correlations. A tolerance level of .40 or less was considered an indication of high multicolinearity.30,31 The tolerance level of each explanatory variable was greater than .50, and the bivariate correlation coefficients were less than .45, indicating that each explanatory variable contributed unique information.

Results

The data for our sample were comparable to reported norms for the explanatory variables, with the exception of BMI and outcome expectations. The prevalence of obesity for African Americans with disability is 36%; in our sample, 55% of the participants were obese.32 The average self-efficacy level in our sample (6.0±2.4) was within 1 standard deviation of normative values (4.92±2) reported for a population with arthritis.20 Average outcome expectations in our sample were 1 standard deviation lower (2.3±0.9) than normative values reported for similarly aged community-dwelling adults with a higher educational level (3.5±0.8).21 The average disability level in our sample (1.6±0.8) was within 2 standard deviations of the disability level (0.77±0.60) reported for a population with arthritis.33 The average perceived social support in our sample (78±18) was within 1 standard deviation of normative values reported for people with chronic health conditions (70±24).28

Notably, 20% of the participants reported that they did not exercise at all, and an additional 6% exercised only 1 time per week. The majority (79%) of these participants with low to no levels of exercise were diagnosed with RA. For those who were exercising, the median and average exercise times were 20 and 29 minutes per day, respectively. The median and average exercise frequency was 3 times per week. The most frequently reported exercises were walking, stretching, and range of motion. The median and average times spent sitting or lying down were 5.5 and 6.2 hours per day, respectively. The median and average times spent engaged in any physical activity were 2.5 and 3.1 hours per day, respectively. The most frequently reported leisure activities were going to church, visiting with family or friends, and babysitting grandchildren. The total physical activity time per week was greater in subjects with RA (X̅=23.75 hours) than in subjects with OA (X̅=16 hours). The results of the t test and chi-square analyses (Tab. 1) revealed no significant differences between subjects with a primary diagnosis of OA and subjects with a primary diagnosis of RA for any of the variables except leisure activity time. Participants with RA spent significantly more time per week engaged in leisure activities.

Table 1.

Characteristics of All Subjects, Subjects With a Primary Diagnosis of Osteoarthritis (OA), and Subjects With a Primary Diagnosis of Rheumatoid Arthritis (RA) and Between-Group Comparison Results

Results of the bivariate linear regression analysis showed that self-efficacy and disability were independently associated with the largest number of physical activities. Age, number of comorbidities, outcome expectations, pain, and social support were not independently associated with any of the physical activity outcomes. The bivariate regression findings are shown in Table 2.

Table 2.

Bivariate Regression Resultsa

An interaction term was included in the multivariate regression equation to assess whether the diagnosis of OA or RA affected the relationship of the explanatory variables to physical activity behavior. A group interaction was present for the following 3 outcome variables: time spent sitting or lying down, leisure activity time, and total physical activity time (Tab. 3). The interaction indicates that the factors associated with the outcome variable were different for subjects with OA and subjects with RA. To further explore the factors associated with physical activity behavior, a multivariate regression analysis was used for subjects with RA. Data for subjects with OA were not analyzed because of the small sample size (n=20). In subjects with RA, none of the variables explained a significant proportion of the variance in exercise time. Time spent sitting or lying down was best explained by disability, pain, and self-efficacy (adjusted R2=.10, P=.05). Leisure activity time was best explained by the number of comorbidities and self-efficacy (adjusted R2=.13, P=.02). Household activity time was best explained by the model containing self-efficacy, outcome expectations, number of comorbidities, and disability (adjusted R2=.18, P=.01). Total physical activity time was best explained by self-efficacy (adjusted R2=.12, P=.01).

Table 3.

Multiple Regression Results

Discussion

Physical activity is considered a critical health behavior, particularly in populations with special needs, such as older adults and people with chronic disease. Physical activity is so critical that the national levels are monitored by the Centers for Disease Control and Prevention through the use of the National Health and Nutrition Examination Survey and the Behavioral Risk Factor Surveillance System. However, despite the importance of physical activity behavior, no gold standard method of measurement exists.34 Additionally, many of the instruments available have been developed for use in non-physically impaired populations and may not adequately capture physical activities performed by populations with physical limitations. For these reasons, comparisons of physical activity levels among different studies are difficult.

In this study, physical activity was defined as “any bodily movement produced by skeletal muscles that results in energy expenditure.”18 The PADS captured the type, frequency, and duration of a variety of different activities in our sample but did not directly capture activity intensity. Indirect evidence of intensity, through the description of reported activities, indicates that we captured physical activities with a very low intensity. The most frequently reported exercise and leisure activities were stretching, range of motion, and socializing. Few subjects reported doing strengthening or aerobic exercises. As with any physical activity measure based on participant recall in a semistructured interview, the accuracy of the data was dependent upon the participant and the interviewer. We used the PADS in a reliable manner, but the extent to which similar findings would have occurred with a different physical activity measure is not known.

The average daily total activity time in primarily nonworking adults was 3.1 hours, indicating that 20.9 hours per day were spent either sleeping, resting, or otherwise being sedentary. The average time spent sitting or lying down, excluding sleep, was greater than 6 hours per day. Comparable sedentary time has been reported for African-American women with physical disabilities.35 The average total physical activity time in age-matched women without musculoskeletal impairments is 13 hours per week more (33.9 hours per week) than the time reported for our sample.36 Our sample of urban African-American women with arthritis demonstrated decreased physical activity time and physical activity intensity levels.

Of the average amount of total physical activity time reported among all subjects, household activities accounted for 46%, leisure activities accounted for 39%, and exercise activities accounted for 15%. Similar proportions of activity types were reported previously for people with RA and for southern rural African-American church members.36,37

In seeking to identify factors that explain physical activity behavior in people with OA and RA, we found that those factors were not the same for the diagnostic groups, despite the similar characteristics of the groups. This finding, to our knowledge, was not reported previously in the literature. Further research is needed for replication and to understand the meaning of this finding.

Self-efficacy was the most consistent explanatory variable for different types of physical activity. Participants with higher confidence for controlling their arthritis symptoms were more likely to spend time engaging in physical activities. In contrast, Eyler and colleagues38 found that for urban Latina and African-American women, exercise self-efficacy either was not associated or was negatively associated with exercise and physical activity behavior. The difference in findings may be attributable to our sample being specific to subjects with arthritis or to the use of different self-efficacy scales.

A limitation of regression analysis is that associations can be found only with variables chosen a priori. Once diagnosis was accounted for, the models explained less than 20% of the variance in the physical activity outcome. The large amount of unexplained variance indicates that physical activity is a complex behavior, and additional variables may elucidate important relationships. Disease duration and previous physical activity behavior have been associated with exercise and physical activity behavior, and inclusion of these variables may have yielded different results.15,16 A limitation of this study is that causal relationships between the outcome and explanatory variables cannot be determined because of the cross-sectional research design.31

Conclusions

This descriptive study provides new information on the physical activity behavior of urban minority women with arthritis. The majority of time was spent engaging in leisure and household activities. The most frequently reported leisure and household activities were the low-intensity activities of socializing and light cleaning. Although the median exercise time was 20 minutes 3 days per week, the most frequently reported exercises were low-intensity stretching, range of motion, and walking.

The factors that best explained variability in physical activity behavior were different in subjects with OA and RA. Self-efficacy was the most consistent explanatory factor. The clinical implications of these findings are that for the goal of increasing physical activity, physical therapists should strive to address all areas of physical activity and develop interventions that relate to leisure and household activities in addition to a traditional home exercise program. Interventions also should include strategies to increase self-efficacy, such as acknowledging the attainment of performance goals and having subjects interact with other subjects with arthritis who have been successful in increasing physical activity.

Footnotes

  • All authors provided concept/idea/research design and consultation (including review of manuscript before submission). Dr Greene, Dr Haldeman, Dr Kaminski, and Dr Neal provided writing and data collection and analysis. Dr Greene provided project management and fund procurement. Dr Lim and Dr Conn provided subjects, facilities/equipment, and institutional liaisons.

  • This study was approved by the Emory University and Grady Healthcare System institutional review boards.

  • Partial funding for this project was received from the Physical Therapy Association of Georgia.

  • The findings were presented at the Combined Sections Meeting of the American Physical Therapy Association; February 23–27, 2005; New Orleans, La.

  • * SPSS Inc, 233 S Wacker Dr, Chicago, IL 60606.

  • Received December 13, 2004.
  • Accepted October 12, 2005.

References

View Abstract