• Physical activity;
  • Exercise;
  • Arthritis;
  • Functional status


  1. Top of page
  2. Abstract


To determine the effect of leisure time and work-related physical activity on changes in physical functioning among 3,554 nationally representative survey respondents, ages 53–63 years in 1994, with arthritis and joint symptoms, interviewed in the Health and Retirement Study (HRS).


In 1992–1994, light and vigorous exercise items were empirically categorized into recommended, insufficient, and inactive leisure time physical activity levels using data from the HRS. Leisure and work-related physical activity levels in 1994 were used to predict 1996 functional decline or improvement, controlling for baseline functional difficulties, health status, sociodemographic characteristics, and behavioral risk factors.


Whereas 29.7% of respondents reported functional declines in 1996, 38.6% of those with baseline difficulties in 1994 reported improvement. Compared with inactive respondents, recommended and insufficient leisure time physical activity were equally protective against functional decline (odds ratio [OR] 0.59 and 0.62, respectively; P < 0.0001). Higher levels of physical activity were also modestly associated with functional improvement among respondents with baseline functional difficulties (OR 1.47, P = 0.05 and OR 1.45, P = 0.01, respectively). Work-related physical activity was not a significant predictor of decline or improvement.


Given the high prevalence of arthritis, even modest increases in rates of lifestyle physical activity among older adults could make a substantial contribution to disability-free life expectancy.


  1. Top of page
  2. Abstract

Despite mounting evidence that physical activity can maintain joint health, prevent disability, and improve function (1–5), the proportion of the US population with arthritis engaging in the recommended amount of physical activity remains low (6, 7). Most health professionals understand the value of exercise (8). However, counseling older adults about how to increase physical activity is challenging in general (9), and becomes particularly complex when arthritis pain becomes a barrier to vigorous exercise. It is therefore of particular interest to examine the extent to which physical activity preserves and protects functional status among older middle-age adults, who experience a high incidence of arthritis symptom onset and overall health decline (10).

This report presents data on individuals with self-reported arthritis and joint symptoms who were enrolled in the Health and Retirement Study (HRS), a nationally representative survey of community-dwelling adults born between 1931 and 1941. Respondents' baseline (1992–1994) average physical activity levels, calculated from HRS questions about leisure time and work-related physical activity, were used to predict change at followup (1994–1996) in self-reported physical functioning. Analyses controlled for the strong baseline correlations between physical activity level and respondents' overall health status, number of chronic medical conditions, preexisting physical functioning difficulties, and other key sociodemographic and behavioral risk factors. HRS data provide a rare opportunity to rigorously examine the question of whether higher levels of baseline physical activity independently protect, or improve, functional status among older middle-age individuals with arthritis.


  1. Top of page
  2. Abstract

Data source.

The HRS is sponsored by the National Institute on Aging and is conducted by the University of Michigan (11). It is a biannual interview study with oversampling of African American, Hispanic, and Florida populations. The current study used publicly available data from 7,946 respondents (81.4% of respondents originally enrolled in 1992 at age 51–61). Patient identifiers had been removed. Complete data on all relevant arthritis status, physical activity, and functional status questions across all of the first 3 (1992–1996) waves of the HRS were used in the study. The study was approved as exempt for human subjects review by the Northwestern University Institutional Review Board.

Study sample and the HRS arthritis case definition.

Arthritis prevalence in 1994 was determined from affirmative responses to 2 questions asked of all HRS respondents: “Have you ever had, or has a doctor ever told you that you have arthritis or rheumatism (bursitis/tendinitis included)?” and “Do you sometimes have pain, stiffness, or swelling in your joints?” Only 2.3% of respondents in 1994 indicated that they had arthritis but did not have joint symptoms. The 3,554 respondents who answered “yes” to both questions were a population-weighted 44.1% of all Americans of the same age, representing ∼8.5 million individuals. The proportion of HRS respondents with arthritis was similar to the 46.2% population-weighted proportion of respondents ages 53–63 years who met the Centers for Disease Control and Prevention arthritis or chronic joint symptoms case definition in the 2001 Behavioral Risk Factor Surveillance System (BRFSS), a telephone interview survey of >215,000 adults that included arthritis questions in every state for the first time in 2001 (12).

Changes in functional status.

Functional status was measured on a 0–9 scale derived from 9 questions asking about difficulties in climbing stairs (1 flight or several), walking (across a room, 1 block, or several blocks), getting in and out of bed, bathing or showering, and eating or dressing without help. Because level of difficulty response options varied slightly across the 1994 and 1996 waves, all responses were dichotomized as some difficulty or none. The Cronbach alpha measure of internal consistency for this 9-item scale in 1994 was 0.81. Because their functional outcomes would primarily be related to surgical treatment, 124 respondents undergoing total joint replacement surgery between 1994 and 1996 were excluded from analyses of changes in functioning.

Although 1994–1996 changes in physical functioning difficulties ranged from −9 (dramatic improvement) to +9 (dramatic decline), the primary analysis addressed the effect of physical activity level on the proportion of respondents with arthritis who experienced any functional decline between 1994 and 1996. Decline was defined as the proportion of respondents reporting a greater number of functional difficulties at followup (1996) than at baseline (1994). A secondary aim was to analyze the effect of physical activity level on improvement in functioning, based on the subpopulation of respondents with at least 1 baseline functional difficulty in 1994 who were thus capable of reporting improvement.

1992–1994 leisure time and work-related physical activity measures.

This study used the 2 original series of HRS “light” and “vigorous” exercise questions, fielded in 1992 and 1994, the only HRS waves with detailed, multiple physical activity questions. These items describe “light physical activity such as walking, dancing, gardening, golfing, or bowling, etc” and “vigorous activities or sports such as aerobics, running, swimming, or bicycling.” In 1992, a separate question was asked about heavy housework, which in 1994 was collapsed into the vigorous activity question. Using previously published methods (13), a single, 0–12 summary measure of total leisure time physical activity was constructed and averaged across the 1992 and 1994 surveys (the correlation between the 1992 and 1994 measures was >0.8). This was done by combining daily, weekly, or monthly exercise frequency responses and weighting vigorous activity frequencies by 2.0 (weighting vigorous activity by 1.5 produced virtually identical results).

This distribution of leisure time physical activity was then trichotomized to approximate frequently used (6, 7) categories that describe progress in meeting the Surgeon General's physical activity intensity and duration recommendations. HRS thresholds for each level were empirically fit to reflect the approximate proportion of respondents with arthritis of the same age at each level of physical activity who were interviewed in the 2001 BRFSS (6). In the BRFSS, the 3 leisure time physical activity categories (tabulated from 6 duration and intensity questions) were “inactive” (<10 minutes of daily moderate or vigorous activity, corresponding to a 0–3 HRS score), “insufficient” (moderate physical activity for <30 minutes a day, or for <3 days a week, or <20 minutes of vigorous activity for <3 days a week, corresponding to a 4–6 HRS score), and “meets recommended level” (at least 30 minutes of moderate activity ≥5 days a week or at least 20 minutes of vigorous activity at least 3 days a week, corresponding to a 7–12 HRS score). Inactive respondents in the 0–3 range of the HRS leisure time activity scale therefore either never engaged in moderate or vigorous activity, or did so less than once a month or only 1–3 times a month. Our analysis of the 2001 BRFSS respondents of the same age (53–63 years) who reported having physician-diagnosed arthritis or chronic joint symptoms demonstrated that 21.8% were inactive, 40.9% were at the insufficient level, and 37.3% reported meeting recommendations. Using the cutoffs listed above for HRS scores, 17.5% of HRS respondents were classified as inactive, 46.8% were classified in the insufficient category, and 35.7% were classified as meeting recommendations in 1994.

The HRS work-related physical activity question was phrased, “My job requires lots of physical effort, such as lifting heavy loads, stooping, kneeling, or crouching.” There were 4 response options: “true almost all of the time,” “most of the time,” “some of the time,” and “none or almost none of the time.” This 0–3 measure was similarly averaged for 1992–1994, based on the assumption that 1992 responses remained the same for those who reported no changes in their job in 1994 (13).

Other baseline characteristics.

A number of other baseline health, behavioral risk factor, and demographic characteristics were used in multivariate analyses, in addition to controlling for the number of functional difficulties at baseline. Potential predictors of differential change in functional difficulties included age, sex, race and ethnicity (African American, English-speaking Hispanic, Spanish-speaking Hispanic, white, or other), household income (3 categories reflecting <300%, 300–499%, and >500% above the 1991 federal poverty level adjusted for household size), years of education (0–8 years, 9–11 years, high school or GED, or >12 years), 1994 current smoking status, 4 body mass index categories in 1994 (underweight [<18.5 kg/m2], normal weight [18.5–24.9 kg/m2], overweight [25.0–29.9 kg/m2], and obese [>30 kg/m2]), 1994 general health status (excellent, very good, good, fair, or poor), and the number of chronic medical conditions reported in 1994. The latter included hypertension, diabetes, heart disease, chronic lung disease, cancer, stroke, or difficulties with vision.

Statistical analysis.

All data are presented as population-weighted proportions using person-level analytic weights. Survey-adjusted chi-square and t-tests were used in bivariate comparisons of respondents with and without arthritis, and comparisons of baseline physical activity levels among respondents with arthritis reporting functional decline or improvement. To test whether physical activity level had an independent effect on 1996 functional status, both linear and logistic regression models were estimated with a priori entry of all baseline covariates, including baseline health and functional status. Coefficients in the linear model of the number (0–9) of 1996 functional difficulties provided estimates of the magnitude of the effect of physical activity, and logistic regression was used to determine the odds of experiencing functional decline and, for those with at least 1 baseline functional difficulty, the odds of reporting functional improvement. All standard errors were adjusted for complex survey design using Stata statistical software version 8 (Stata, College Station, TX).


  1. Top of page
  2. Abstract

Baseline characteristics of the arthritis population.

Baseline descriptive statistics for HRS respondents are presented in Table 1, with complete 1992–1996 data comparing all those who reported having arthritis and joint symptoms in 1994 with all other HRS respondents. The results show that respondents with arthritis and joint symptoms were significantly older; more likely to be female; and included a higher proportion of African American, lower income, and less educated respondents. As expected (10), respondents with arthritis were much more likely to be obese, to report a greater number of chronic medical conditions, and to have dramatically lower overall health status.

Table 1. Comparative baseline characteristics of health and retirement study respondents reporting arthritis and joint symptoms in 1994*
CharacteristicsNo arthritis or joint symptoms (n = 4,392)Arthritis and joint symptoms (n = 3,554)
  • *

    Unless otherwise indicated, values are the percentage. GED = General Education Development test; BMI = body mass index.

  • Includes 82 respondents who indicated they had arthritis but did not “sometimes have pain, stiffness, or swelling in your joints.”

  • Includes 124 respondents who subsequently underwent total joint replacement surgery by 1996.

  • §

    P < 0.01, standard errors for all proportions ≤2%.

All respondents55.944.1
Mean (SE) age, years§55.7 (0.05)56.3 (0.06)
 White/other race86.184.2
 African American§8.510.3
 Hispanic (Spanish-speaking)5.25.3
 Hispanic (English-speaking)3.33.1
 0–8 years§7.711.1
 9–11 years§11.717.4
 High school diploma or GED35.238.6
 >12 years§45.332.7
Household income as a percent of the 1991 Federal poverty level adjusted for household size  
Current smoker25.826.9
 Underweight (BMI < 18.5 kg/m2)1.21.1
 Normal (BMI 18.5–24.9 kg/m2)§39.731.3
 Overweight (BMI 25.0–29.9 kg/m2)41.739.8
 Obese (BMI > 30.0 kg/m2)§17.427.8
Number of chronic medical conditions 1994, mean (SE)§0.72 (0.01)2.1 (0.02)
Health status in 1994§  
 Very good34.827.5

Association of baseline physical activity and functional status.

In 1994, only 11.6% of HRS respondents with arthritis reported having difficulty in walking across a room, dressing, eating, getting in and out of bed, or bathing or showering without help. However, at least one difficulty with climbing stairs or walking was frequently reported (49.0%). When combining the 9-item scale, 50.1% of respondents with arthritis had no baseline functional difficulties, 23.0% reported a single functional difficulty (almost always difficulty walking several blocks or climbing several flights of stairs), and 26.9% reported ≥2 difficulties, including 4.2% who reported ≥5 functional difficulties.

As expected, the association between 1992–1994 leisure time physical activity (3 levels) and the number (0–9) of 1994 functional difficulties was strong (r = 0.49, P < 0.0001). This is illustrated in Figure 1, which displays the distribution of HRS respondents at 3 levels of baseline functional status. Only 7.4% of all respondents without any functional difficulties were inactive, whereas 37.4% of all respondents with ≥2 functional difficulties reported being inactive (P < 0.0001).

thumbnail image

Figure 1. Estimated US population ages 53–63 with arthritis and joint symptoms by number of baseline functional difficulties and 1992–1994 average leisure time physical activity levels.

Download figure to PowerPoint

There were large baseline differences in the mean number of physical difficulties at each leisure time physical activity level (0.52 at the recommended level, 1.09 at the insufficient level, 2.59 at the inactive level; P < 0.0001 for all comparisons). However, respondents reporting any level of physical activity at work had virtually identical mean baseline functional difficulties (between 0.72 and 0.78), which was significantly better compared with respondents with no work-related activity, who had a mean of 1.54 functional difficulties (P < 0.0001). Therefore, in subsequent analyses of work-related physical activity, those with any level of work-related activity were combined into a single group and compared with respondents who were inactive at work or not working.

Physical activity and transitions in physical functioning.

With respect to 1994–1996 changes in functioning, 29.7% of all respondents with arthritis reported functional declines (additional functional difficulties) in 1996. Of those who reported functional decline, 28.3% reported no baseline functional difficulty in 1994, 34.6% reported only 1 difficulty, and 28.1% reported ≥2 previous difficulties. When only those 1,742 respondents capable of improving were considered (i.e., excluding slightly more than half of all respondents who had no functional difficulties at baseline in 1994), 38.6% reported fewer 1996 difficulties.

Figure 2 displays the proportion of respondents at each leisure time physical activity level reporting functional declines, and functional improvement for those with baseline difficulties, at each level of baseline function (no difficulties, only 1 difficulty, ≥2 difficulties, respectively). Figure 2 illustrates the protective effect of higher levels of activity on respondents with none or 1 baseline difficulty, who accounted for approximately three-quarters of the sample. Although an association between increasing activity and functional improvement was seen in respondents with 1 baseline difficulty, the gradient is less clear for those with ≥2 baseline functional difficulties. For that group, inactive respondents still had the highest rate of decline, but improvement was not associated with increased physical activity. When results were combined for all respondents, 27% of those at the recommended physical activity level and 29% of those at the insufficient level experienced declines, compared with 37% of respondents at the inactive level (P = 0.0004). There was no significant difference between work-related physical activity levels; 28.6% of those active at work experienced additional difficulties as compared with 30.9% of those inactive at work or not working (P = 0.18).

thumbnail image

Figure 2. 1994–1996 changes in physical functioning among respondents at each of 3 levels of baseline physical functioning, by level of baseline leisure time physical activity. A, No baseline functional difficulties in 1994; B, One baseline functional difficulty in 1994; C, Two or more baseline difficulties in 1994. Black = improved, grey = stayed the same, white= declined.

Download figure to PowerPoint

Multivariate analyses of functional decline or improvement.

Both linear (R2 = 0.53) and logistic regression models found that, as compared with inactive respondents, respondents with arthritis who engaged in either recommended or insufficient leisure time physical activity had significantly better 1996 functional outcomes. Both models obtained very similar estimates of the effects of both physical activity and baseline covariates; therefore, for ease of interpretation, Table 2 provides logistic regression results only. The odds ratios (ORs) presented indicate whether a factor conferred risk (>1.0) or was protective (<1.0) against functional decline between 1994 and 1996.

Table 2. Logistic regression results predicting the likelihood of functional decline in 1996 in population-weighted respondents with arthritis and joint symptoms in 1994 (n = 3,430)*
 Odds ratio (95% CI)P
  • *

    Excludes 124 respondents who underwent total joint replacement 1994–1996. 95% CI = 95% confidence interval; GED = General Education Development test; BMI = body mass index.

Leisure time physical activity  
 Recommended0.59 (0.44–0.78)< 0.0001
 Insufficient0.62 (0.48–0.79)< 0.0001
Work-related physical activity0.97 (0.90–1.06)0.60
Age, years0.97 (0.95–0.1.00)0.05
Female1.68 (1.36–2.06)< 0.0001
 African American0.97 (0.80–1.18)0.80
 Hispanic (Spanish-speaking)0.68 (0.42–1.10)0.12
 Hispanic (English-speaking)1.09 (0.71–1.69)0.66
 0–8 years1.82 (1.33–2.49)< 0.0001
 9–11 years0.18 (0.04–0.31)0.009
 High school diploma or GED1.35 (1.04–1.73)0.02
 >12 yearsReference 
Household income as a percent of the 1991 federal poverty level adjusted for household size  
 <300%1.20 (0.96–1.51)0.09
 300–499%0.97 (0.73–1.29)0.84
Current smoker1.73 (1.42–2.10)< 0.0001
 Underweight (BMI < 18.5 kg/m2)2.96 (1.46–5.98)0.003
 Overweight (BMI 25.0–29.9 kg/m2)1.53 (1.23–1.91)< 0.0001
 Obese (BMI > 30.0 kg/m2)2.12 (1.67–2.69)< 0.0001
 Normal (BMI 18.5–24.9 kg/m2)Reference 
1994 functional status (0–9 difficulties)0.71 (0.65–0.76)< 0.0001
Number of chronic medical conditions1.17 (1.07–1.28)0.001
Health status  
 Excellent0.34 (0.20–0.55)< 0.0001
 Very good0.49 (0.32–0.75)0.002
 Good0.62 (−0.65–0.15)0.02
 Fair0.85 (0.59–1.21)0.37

As compared with inactive respondents, those at the recommended and insufficient activity levels had a 59% and 62% likelihood of decline in 1996, respectively (both comparisons P < 0.0001). Results from the linear model (data not shown) indicate that recommended activity was associated with 0.35 fewer 1996 difficulties and insufficient activity was associated with 0.33 fewer 1996 difficulties as compared with inactive respondents (P < 0.0001). However, work-related activity was only slightly protective and nonsignificant. As expected, baseline functional difficulties explained most of the variance in both models. The protective effect of additional baseline difficulties in Table 2 reflects the fact that many respondents with none or 1 baseline difficulty in 1994 experienced additional 1996 functional difficulties, whereas many with more than 1 baseline difficulty reported improvement in 1996 (Figure 2).

Although age, race, ethnicity, and income level differences were modest, female sex (OR 1.68, P < 0.0001) was a highly significant risk factor for functional decline. When compared with college-educated respondents, those with a high school education or less had significantly higher rates of functional decline. Being a smoker was associated with 1.73 more 1996 functional difficulties (P = 0.0001). As compared with normal weight respondents, being underweight, overweight, or obese was associated with a significantly elevated likelihood of additional functional difficulties. Each additional 1994 chronic condition was associated with 1.17 times the likelihood of reporting additional functional difficulties (P = 0.001) in 1996. Compared with poor overall health at baseline, excellent, very good, and good overall health were highly protective against functional decline, although fair health was not significantly different from poor health.

The same logistic regression model was run for respondents with any baseline functional difficulties to predict functional improvement (versus staying the same or declining). Compared with inactive respondents, those at the recommended level of leisure time activity had 1.47 times the likelihood of improvement (P = 0.05), and those at the insufficient level had 1.44 times the likelihood of improvement (P = 0.01). Once again, work-related activity was nonsignificant (data not shown).


  1. Top of page
  2. Abstract

Among older middle-aged adults, the most important predictors of future health and functioning are baseline health and functioning (14–16), and these results further document the widely observed correlation between baseline health status, functioning, socioeconomic status, and health outcomes (17). It is therefore remarkable that the results nevertheless provide consistent evidence of the protective effects of lifestyle physical activity on functional decline after controlling for those factors.

Although some researchers have argued for the importance of occupational physical activity for health outcomes (18), there were only weak associations between functional changes and the HRS single-item measure of any work-related physical activity. This finding may reflect the fact that only slightly more than half (56%) of the HRS respondents with arthritis were even employed for pay in 1994, and only 15% reported the highest level of physical activity at work. The fact that there were virtually no differences in functional outcomes between levels of work-related activity indicates that this may have been a poorly specified measure in this study. It is also possible that respondents with no work-related physical activity may have had relatively higher levels of leisure time activity, thus further undermining any independent effects of work activity on functioning.

A key limitation of this and all other observational research studies of physical activity is the potential for residual confounding. It is likely that higher levels of physical activity reflect other, unmeasured health-seeking behaviors, and that individuals who exercise regularly may have other characteristics that contribute to preserving functioning. Although the HRS provides a large sample size and a uniquely extensive set of control variables, the findings for exercise reported here are hypothesis generating and await validation in sufficiently rigorous controlled trials.

In addition to problems with survey attrition and nonspecific arthritis prevalence estimates, the HRS physical activity questions provide a relatively crude approximation of physical activity intensity. This and the fact that self-reported activity levels are frequently exaggerated when compared with objective measures (19) may account for the relatively similar findings for both the insufficient and recommended activity levels. Linear regression results using the continuous HRS physical activity scale indicate that each additional HRS activity point at baseline was associated with approximately one-third fewer physical functioning difficulties at followup (P = 0.003). Although it was difficult to quantify physical activity intensity thresholds with HRS data, the results nevertheless imply that physical inactivity is an important functional risk factor compared with even modest activity levels.

The extent to which arthritis specifically impacted functional declines reported in the HRS is unknown. Ideally, it would have been of interest to model the effect of physical activity on arthritis-specific symptoms such as joint pain. However, the HRS provides only very crude measures of symptom status that do not account for differences between treated and untreated pain. The contribution of arthritis progression to overall functional decline is a complex process that requires significant additional research.

Finally, the level of difficulty or need for assistance were not uniformly collected in the HRS functional mobility and facility items. In particular, the amount of functional improvement frequently reported among older survey respondents probably reflects significant “noise,” or measurement error related to panel conditioning and response variability (changes in the meaning of the question) (20). Exaggeration of self-reported functional improvement may account for the relatively weak predictive performance of models of functional improvement as compared with models of functional decline.

The data from the HRS presented here add to a growing literature that suggests that physical activity is important for maintaining functional ability in individuals with arthritis. At the population level, Dunlop and colleagues have shown that elderly adults with arthritis participating in the Asset and Health Dynamics of the Oldest Old Survey cohort who reported engaging in vigorous physical activity were at lower risk of functional decline (OR 0.64 after 3 years of followup) (21). A clinical study by Sharma and colleagues found that persons with knee osteoarthritis who reported at least 60 minutes of aerobic exercise per week were at lower risk of having poor arthritis-related function (OR 0.75) after 3 years of followup (22). There are also a number of controlled trials that support the participation of patients with arthritis in formal exercise programs, which have been found to increase strength, decrease pain and fatigue, and improve quality of life without adversely affecting joint status (23–25).

Among older Americans who report having difficulty walking or shopping, arthritis is frequently the leading cause of their mobility difficulty, and many experience costly treatment for disabling chronic joint symptoms (26–28). However, among the elderly population in general, national health surveys document remarkable age-adjusted improvements in self-reported functional status of older Americans (29–32). One modeling projection (31) indicates that the doubling of total joint replacement operations over the 1980s and 1990s may, in itself, account for 50% of declining disability rates over this same period. Increasing rates of lifestyle physical activity among older adults may make an even greater future contribution to disability-free life expectancy.


  1. Top of page
  2. Abstract