Assessing Physical Activity in Persons with Rheumatoid Arthritis Using Accelerometry : Medicine & Science in Sports & Exercise

Journal Logo


Assessing Physical Activity in Persons with Rheumatoid Arthritis Using Accelerometry


Author Information
Medicine & Science in Sports & Exercise 42(8):p 1493-1501, August 2010. | DOI: 10.1249/MSS.0b013e3181cfc9da
  • Free


The objective measurement of physical activity (PA) using accelerometry is now used in large-scale epidemiologic studies (22) and clinical trials including studies of persons with arthritis (7,10,16). Rheumatoid arthritis (RA) affects approximately 1% of adults in the United States, and it has been recognized as one of the most critical rheumatologic conditions in the developed world (20), largely because of the associated joint damage, declines in functional status, and premature mortality (15,19). The disease process leads to chronic and progressive inflammation involving multiple joints as well as other organ systems. RA is associated with substantial disability, economic losses (6), and even reduced life expectancy as a result of disease consequences (17). PA has been found to reduce pain and improve function in persons with RA, hence the recent interest in measuring PA in this population. An important advantage of objective PA measurement by accelerometers is that they capture salient behavior characteristics including the frequency, intensity, and duration of PA, as well as periods of rest. Recently, waist-mounted accelerometers were used in population surveillance of PA in the National Health and Nutrition Examination Survey (NHANES), indicating that they can be used in population and clinical studies.

An important analytical step is the translation of accelerometer counts into PA measurements. Numerous validation studies have focused on interpreting accelerometer counts by defining thresholds that correspond to light-, moderate-, and vigorous-intensity physical activities (8,21). In contrast, few have been published on data issues related to extracting meaningful PA descriptions from accelerometer output. Masse et al. (12) coined the term "accelerometer data reduction" for the software process that translates accelerometer counts into physical measurements of activity intensity, frequency, and duration. A basic building block in this process is the assessment and interpretation of "nonwear time." Nonwear relates to periods when activity counts on the accelerometer readout register as "0" because the monitor is not being worn by the subject. Time spent in sedentary activities, which can also register as "0," is often indistinguishable from nonwear in accelerometer readouts. Thus, some upfront decisions must be made regarding the maximum number of minutes of consecutive "0" counts that will be allowed in the data before considering those minutes of "nonwear." Nonwear time is a structural foundation in the data reduction process. Using a predetermined threshold for the number of hours that constitute a "valid day" of accelerometer wear, nonwear time becomes the key variable in deciding which days "rule in." Thus, how one defines "nonwear" has major ramifications in data reduction. The influence of these arbitrary procedural decisions on the assessment of PA outcomes was illustrated by Masse et al. (12). To our knowledge, these parameter definitions for accelerometer data reduction have not been examined for persons with arthritis.

A landmark PA study in the general US adult uninstitutionalized population was based on accelerometer measurements from the NHANES published by Troiano et al. (22). An important contribution of that study was the report of their accelerometer data parameter definitions of 1) nonwear time and 2) a valid day of monitoring (determined from unpublished internal empirical analyses (personal communication, October 18, 2008)). This information provided a valuable starting point for other studies that require data reduction from accelerometer output. However, parameter value definitions from this benchmark study of the general US population may not be applicable to subpopulations with chronic conditions. For example, PA patterns in persons with arthritis may differ from the general population because of pain and stiffness that necessitate frequent extended rest breaks.

The purposes of this study were to investigate empirically if data reduction parameter value definitions derived from the general adult US population were appropriate to apply to accelerometer data from persons with RA for the purpose of evaluating PA outcomes and to explore the impact of applying alternate parameter definitions to that accelerometer data.


Study population and sample.

This study analyzed data from 107 persons with RA who participated in the baseline (2006-2008) accelerometer assessment from two concurrent studies with common inclusion/exclusion criteria: an ongoing randomized clinical trial, Increasing Motivation for Physical Activity in Arthritis Clinical Trial (IMPAACT), and a second study of lifestyle physical activity measurement in adults with arthritis (LPAM). The IMPAACT study is an ongoing randomized trial of lifestyle PA promotion. Participants were randomized to receive either the IMPAACT intervention, a tailored PA counseling program in addition to a brief physician encounter promoting PA, or a control intervention consisting only of the physician advice to increase PA. Follow-up measures include repeated accelerometer, health status, and physical function assessments every few months for 2 yr. The LPAM study examined the relationship among 7-d accelerometry, physiologic fitness parameters, and a measure of self-reported PA. Both studies received IRB approval at Northwestern University, and written informed consent was obtained from each of the participating subjects. Inclusion criteria for RA participants were as follows: 1) aged 18 yr or older, 2) satisfy American College of Rheumatology criteria for RA (1,2), 3) able to ambulate at least 50 ft, and 4) speaks English. Exclusion criteria were as follows: 1) planned total joint replacement in the subsequent 12 months; 2) contraindication to PA because of comorbid condition including a history of peripheral vascular disease, spinal stenosis, residual lower extremity neuromuscular effects of stroke, and major signs or symptoms suggestive of pulmonary and cardiovascular disease; 3) unable to perform basic self-care activities; or 4) plans to relocate away from the Chicago area within 24 months. No instructions were given for daily activity or exercise before accelerometer measures were obtained in these preintervention assessments.

Accelerometer measures and procedures.

PA was monitored in all study participants using a GT1M ActiGraph accelerometer. The GT1M ActiGraph (Pensacola, FL) is a small uniaxial accelerometer that measures vertical acceleration and deceleration (13). The acceleration signal is filtered and digitized by an eight-bit analog-digital (A-D) converter at 30 samples per second. The A-D converter measures the magnitudes of the accelerations. Accelerometer output displays as activity counts (vertical movements per 1-min epochs). Spurious accelerometer counts were identified by negative counts; these spurious values were set to missing on a minute-by-minute basis. The validity and reliability of ActiGraph accelerometers under field conditions have been established in many populations including RA (3,9,14,23).

Participants were instructed to wear the accelerometer upon arising in the morning and to wear it continuously (except for water activities) until going to bed at night for seven consecutive days. The unit was worn on a belt at the natural waistline on the right side of the hip in line with the right axilla (Fig. 1). Participants also maintained a daily log (time sheet) to record when the accelerometer was put on in the morning and removed at night. Skipped days reported on the time sheet were excluded from the analysis.

Placement of the waist-mounted accelerometer.

Accelerometer data reduction parameter definitions.

Nonwear refers to a sustained period showing little or no activity that may represent an interruption in accelerometer monitoring. A benchmark definition for nonwear applied by Troiano et al. (22) to US adult accelerometer data consisted of an interval of at least 60 min of zero activity counts that contained no more than 2 min of counts between 0 and 100. In this RA sample, we examined candidate nonwear periods ranging from 20 to 300 min of zero activity count. A rolling window algorithm (i.e., started from the beginning of each day, scanned each minute, and began a potential nonwear period when a zero activity count was found) was used to identify nonwear periods. Consistent with the approach of Troiano et al. (22), a nonwear period ended either when a third minute of activity counts fell in the range of 0-100 or when a 1-min activity count was greater than 100 (18). Although nonwear has been frequently indistinguishable from sedentary activity, for the purposes of this article, the "nonwear" terminology of Masse et al. (12) is used throughout. The definition of a valid day was based on wear hours that exceed a threshold. The benchmark valid day definition from Troiano et al. (22) for US adults was a day containing 10 h or more of wear time, defined as 24 h minus nonwear time. Most importantly, the 10 h of wear time did not need to be continuous to be considered valid.

PA outcomes

The effects of selected parameter definitions were examined in regard to PA outcomes. The amount of PA was evaluated for each person in four ways: 1) mean daily activity counts, 2) activity counts per wear hour, 3) mean daily minutes of moderate to vigorous PA (MVPA) according to count thresholds that occur in 10-min bouts, and 4) MVPA bout minutes per wear hour. Mean daily activity counts represented the summed activity counts for all wear hours divided by the number of days monitored. Activity counts per wear hour was the ratio of total activity counts from the wearing periods to the total wear hours during the 7-d monitoring period. Mean daily minutes of MVPA (occurring during bouts) were obtained by applying count thresholds corresponding to moderate or vigorous PA summed over all wear hours divided by the number of monitored days for each person. This outcome is often relevant for the assessment of activity relative to meeting public health guidelines. MVPA bout minutes per wear hour were calculated as the ratio of total minutes of MVPA (occurring during bouts) to the total wear hours during the 7-d monitoring period. For analytic purposes in this article, we used the 10-min activity bout definition applied by Troiano et al. (22). A "bout" was then defined as the activity minutes that occurred during a period of at least 10 consecutive minutes of activity above the moderate/vigorous threshold of 2020 counts, allowing for up to 2 min of interruptions below that threshold during the bout. A potential bout was therefore begun with an activity level above 2020 counts per minute and was terminated when 3 min of activity counts occurred below that threshold. Only potential bouts that lasted more than 10 min were included in this definition of bouts.


Descriptive analyses graphically depicted the relationship between nonwear parameter definitions and daily wear outcomes: daily wear hours and daily nonzero minutes. This process was used to identify candidate nonwear parameter values that were associated with stable results in regard to daily wear outcomes when applied to accelerometer data from RA patients.

We compared candidate nonwear parameter definitions applied to the RA data with the 60-min nonwear benchmark of Troiano et al. based on the US adult population on the following variables: mean daily activity counts, activity counts per wear hour, mean daily MVPA bout minutes, and MVPA bout minutes per wear hour. Statistical testing using Student's t-test evaluated within-person paired differences in PA outcomes related to selected wear time parameter values (e.g., 90 or 120 min). Specifically, nonparametric comparisons of PA outcomes resulting from the application of candidate and benchmark parameter values were performed using quantile regression to estimate the median difference and SE because of skewed distributions. The related Student's t-statistic on the basis of the ratio of the median difference to its SE was used to maintain overall testing at an α = 0.05 level of testing (11). Student's t-statistic mirrors a t-test, correcting for experiment-wise error rate to account for multiple comparisons. A weighted κ coefficient was used to describe the agreement in the estimated number of valid days resulting from the application of different parameter values. A κ coefficient is a statistical index that compares the agreement against that which might be expected by chance and ranges from +1 (perfect agreement) to −1 (complete disagreement). A weighted κ extends a κ statistic to ordinal data; it assigns less weight to categories that are farther apart (5). We used Stata/SE 10.0 for quantile regression. Other analyses were performed using SAS software version 9.2 (Cary, NC).


A total of 107 participants meeting American College of Rheumatology criteria for RA participated in accelerometer monitoring. The age distribution of the sample was broad (range = 23-80 yr, mean = 54 yr) but skewed to older ages. As expected, subjects were predominantly female (85%) and college-educated (64%), and most of the subjects were white (76%). On average, participants had RA disease duration of 13.8 yr (SD = 11 yr). Overall, 35% of the participants reported having comorbidities that may have affected their ability to be mobile. All 38 participants with comorbidities reported having secondary osteoarthritis; additional comorbidities included back problems (n = 37), respiratory conditions (n = 35), cardiovascular conditions (n = 33), depression (n = 20), and osteoporosis (n = 20). Accelerometer assessment used 7 d of continuous monitoring during waking hours using a GT1M ActiGraph motion sensor (except for skipped days noted on the log, which were removed from analyses). These 107 RA patients collectively contributed 737 d of accelerometer data.

Nonwear represents interruptions in accelerometer data collection. However, it is important to note that both nonwear and very sedentary activities register as "0" counts in accelerometer readouts, requiring investigators to develop an arbitrary threshold for defining nonwear. In this study, participants were permitted to remove the units if they took a nap, bath, or shower during the day. On average, units were removed 1.2 times per day, which included taking the unit off at bedtime (calculated from on/off times reported by participants via logs). Candidate nonwear definitions, ranging from thresholds of 20-300 consecutive minutes of zero activity counts, allowing for up to 2 min of very light activity (<100 activity counts), were investigated to identify potential interruptions in wear time as part of the data reduction process.

Graphs show the relationship between candidate nonwear definitions and the following daily wear outcomes: mean daily wear hours (Fig. 2A), and mean daily nonzero activity count minutes (Fig. 2B). As expected, Figure 2A shows that the average daily wear hours increased with length of maximum continuous minutes of zero counts allowed (labeled as "nonwear threshold"). Figure 2B shows that the mean number of nonzero activity count minute levels increased with the nonwear parameter value until it stabilized at 478 min·d−1 of activity, which corresponded to the 90-min maximum continuous minutes nonwear definition.

A, Mean daily wear hours by threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring. B, Mean daily activity minutes (nonzero counts) by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring.

The influence of nonwear parameter values on PA outcomes was investigated graphically for activity counts in Figure 3A and MVPA bout minutes in Figure 3B. Activity counts presented in Figure 3A were evaluated in mean daily activity counts and activity counts per wear hour. The mean daily activity counts shown in Figure 3A increased as the nonwear parameter value increased (labeled "nonwear threshold") but stabilized near the 90- to 120-min maximum continuous minutes nonwear definition. In contrast, activity counts per wear hour did not stabilize but consistently decreased as nonwear parameter values increased. Minutes of MVPA (occurring during bouts), presented in Figure 3B, were evaluated in mean of daily MVPA bout minutes and MVPA bout minutes per wear hour. It was evident that the mean of daily MVPA bout minutes was invariant to the nonwear parameter values. This invariance was expected because the nonwear period was terminated by definition when a minute of MVPA occurred. As a result, the same amount of MVPA minutes occurring in bouts during a day was captured by all nonwear parameter values. In contrast, the MVPA bout minutes per wear hour decreased as the nonwear parameter values increased because the numerator was stable, but the denominator was increasing.

A, Accelerometer activity counts by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring. B, Minutes of moderate/vigorous activity bouts by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring.

Summaries of PA outcomes resulting from the application of nonwear parameter values of the benchmark 60-min threshold and candidate 90- and 120-min thresholds are shown in Table 1. There were significant differences between paired comparisons of average daily activity counts captured by the 60-min nonwear parameter value compared with each of the 90- and 120-min values on the basis of Student's t-tests, but these differences were very small (<0.1%). There was no significant difference in average daily total activity counts between the 90- versus the 120-min nonwear parameter values, consistent with Figure 3A that shows stabilization at these values. Table 1 shows significant differences related to activity counts per wear hour for all contrasts; these differences were moderate in magnitude, ranging from 5% to 7%. Statistical testing was precluded for MVPA minutes, which was identical for all parameter values (i.e., variance = 0). No statistical differences were demonstrated for pairwise contrasts of MVPA bout minutes per wear hour, and the magnitude of differences was small.

Physical activity outcomes by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring.

A valid day of wear was defined on the basis of the number of wear hours that exceeded a predetermined arbitrary threshold. Figure 4 shows the proportion of data retained as a function of the definition of a valid day and the nonwear parameter definition. For example, at the 60-min nonwear parameter benchmark, the 10-h daily wear minimum resulted in 11% data loss (i.e., corresponding to 89% data retention). Increasing the daily wear minimum requirement to 12 h more than doubled data loss to 24%. The curves from this RA sample related to the 8- and 10-h valid day thresholds stabilized and captured more than 90% of monitored days near the 90- to 120-min nonwear parameter values. In contrast, a 12-h wear threshold required much larger nonwear threshold values to capture more than 90% of monitored days. These graphs suggest that the benchmark threshold of 10 wear hours or more used in the general US adult population to identify a valid day is applicable in the context of RA, even when nonwear parameter values of 90 or 120 min are used.

Percentage of monitored days with daily wear hours exceeding 8, 10, and 12 h by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring.

Agreement in the distribution of valid days for this RA sample was evaluated using weighted κ coefficients for nonwear parameter values of 60, 90, and 120 min in Table 2. Consistent with Figure 4, the frequency of valid days was greatest for the 120-min nonwear parameter definition, followed by the 90- and 60-min definitions. The agreement in the number of valid days was very strong for the 90- and 120-min nonwear parameter values (weighted κ = 0.92). Agreement between the 60-min nonwear parameter value and the higher parameter values was moderate (0.67-0.75).

Distribution of valid days by nonwear threshold minutes from 107 persons with RA contributing 737 d of accelerometer monitoring.


This article describes accelerometer methodology applied to the special population of persons with RA. There are important reasons to strive for the most accurate measure of PA possible in persons with potential movement limitations. First and foremost, those with movement limitations are at high risk for becoming disabled over time, adding tremendous financial burden to health care costs as the population ages. It is known that persons with arthritis report PA levels below that of the general population (4). Appropriate PA programs cannot be formulated for targeting this segment without a clear picture of baseline activity levels and patterns. Similarly, evaluating the effectiveness of public health programs that seek to increase PA in any population will be difficult without adequate instrumentation and interpretive technology. For this study, a uniaxial accelerometer was chosen over a triaxial model such as the Stay Healthy RT3 because the uniaxial was less expensive at the time units were purchased for this study and had been used in other large epidemiologic studies. No one instrument is preferable in persons with RA.

Using data reduction parameters honed on a sample of the general population is especially problematic in persons with RA, who may be more sedentary secondary to joint pain and stiffness and have fatigue issues that require longer rest breaks. Indeed, Figure 2B indicates that for persons with RA, alternate definitions of nonwear that allow a maximum of 90 min of zero activity may be more appropriate than the nonwear definition based on 60 min of allowable zero activity applied to the NHANES sample of general population adults. This finding supports the revision of data reduction parameters for those with RA that exceed the 60-min rule of Troiano et al. (22). The result of this adjustment is that, although allowing for these longer periods of rest does not change the final data very much, it does allow the retention of more valid measurement days in this population.

It is also interesting to note that descriptions of PA outcomes shown in Table 1 are similar for the 60-, 90-, and 120-min nonwear parameter definitions. This finding might be explained by considering the person who takes a rest break to nap or watch a movie, activities that would not be registering accelerometer counts; their overall activity for the day would probably remain the same, whether the nap or the movie lasted 90 or 120 min, hence the similarity in MVPA bout minutes (both tabled entries). Data regarding employment were not collected for this study, but it is feasible that a similar pattern would emerge for persons with sedentary office jobs who must sit for long periods during meetings or computer work.

Possibly the most intriguing finding of the analyses was that total activity counts and MVPA bouts seemed to be more robust PA descriptors than ratios on the basis of wear hours. This is especially critical if one's goal is to compare PA behaviors across studies using accelerometry. What might account for this finding? One must consider potential sources of "0" counts reading out during accelerometry wear. Although one is tempted to write off "0" readings to nonwear (thus reducing the total amount of valid wear hours), subjects may in fact be sitting very quietly, as one might during prolonged periods of sitting that involve reading, watching TV, or computer activities. Thus, perhaps a better term to use is "nonmovement time" rather than assume that the presence of zeros reflects nonwear time. Indeed, in a study involving persons sitting quietly for a resting metabolic rate before activity while simultaneously wearing an accelerometer, all corresponding accelerometer readouts for that period registered as "0" counts (unpublished data). Sitting activities, already prevalent in the general population, may be especially attractive to those with painful or stiff joints. In fact, any person who may have received physical or occupational therapy will have been encouraged to take frequent rest breaks to accommodate their symptomatic joints.

While extending the allowed limit on continuous zero count minutes in those with arthritis seems reasonable, the "valid day" definition based on the 10-h rule of Troiano et al. (22) seems to be applicable for this population. Accepting the 10-h daily wear minimum resulted in 11% data loss (i.e., corresponding to 89% data retention), while increasing the daily wear minimum to 12 h more than doubled data loss to 24%. It may be impractical to expect a 12-h activity day from persons with a condition that includes fatigue and joint pain as their most prominent symptoms. The curves from this RA sample related to the 8- and 10-h valid day thresholds stabilized and allowed for almost 90% data retention, but insisting on a minimum of 12 h of wear more than doubled data loss. Because the average wear time for the accelerometers in this sample was 14 h, it is reasonable that the 10-h rule would include 93% of the study sample (Fig. 4). However, in samples with less compliance to wearing the accelerometer during all waking hours, the 10-h rule may cause excessive elimination of subjects from data analyses.

Finally, in performing these analyses, it is evident that the plea for more attention to methodology disclosure in studies using accelerometer data by Masse et al. (12) is well founded because variability may exist from population to population and may alter data interpretation considerably. Not having exact information about how Troiano et al. (22) accomplished their data reduction made comparison of methods impossible, but having information about the parameters used in their rigorous study of the general US population was extremely useful, nonetheless, because no benchmark on a large sample had been established.

In summary, this study examined data reduction techniques applied to accelerometry output in persons with RA compared with parameters applied to the general public. Because of the difficulty of attributing bouts of continuous zeros to nonwear versus inactivity, we suggest referring to bouts of continuous zeros as nonmovement time. On the basis of the findings from this study, we argue for increasing the allowed nonmovement time bouts of continuous zeros in this RA subpopulation from 60 to 90 min, while retaining the 10-h day as the measure of the "valid day."

The authors gratefully acknowledge the National Institute of Nursing Research (grant No. P30 NR009014, Center for Reducing Risks in Vulnerable Populations) and the National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01 AR052912-01, R01 AR055287, R01 AR054155, P60 AR048098) for their support in completing this work.

The authors also gratefully acknowledge the insightful suggestions of Dr. David Berrigan.

The results of the present study do not constitute endorsement by American College of Sports Medicine.


1. Altman R, Asch E, Bloch D, et al. Development of criteria for the classification and reporting of osteoarthritis. Classification of osteoarthritis of the knee. Diagnostic and Therapeutic Criteria Committee of the American Rheumatism Association. Arthritis Rheum. 1986;29:1039-49.
2. Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum. 1988;31:315-24.
3. Brage S, Wedderkopp N, Franks PW, Andersen LB, Froberg K. Reexamination of validity and reliability of the CSA monitor in walking and running. Med Sci Sports Exerc. 2003;35(8):1447-54.
4. Centers of Disease Control and Prevention. Arthritis [Internet]. [cited 2009 May 12]. Available from:
5. Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull. 1968;70:213-20.
6. Committee AH. Guidelines for the management of rheumatoid arthritis. American College of Rheumatology Ad Hoc Committee on Clinical Guidelines. Arthritis Rheum. 1996;39(5):713-22.
7. Farr JN, Going SB, Lohman TG, et al. Physical activity levels in patients with early knee osteoarthritis measured by accelerometry. Arthritis Rheum. 2008;59(9):1229-36.
8. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc. 1998;30(5):777-81.
9. Hendelman D, Miller K, Baggett C, Debold E, Freedson P. Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med Sci Sports Exerc. 2000;32(9 suppl):S442-9.
10. Hirata S, Ono R, Yamada M, et al. Ambulatory physical activity, disease severity, and employment status in adult women with osteoarthritis of the hip. J Rheumatol. 2006;33(5):939-45.
11. Hochberg Y, Tamhane AC. Multiple Comparison Procedures. New York (NY): John Wiley & Sons, Inc; 1987. p. 80-5.
12. Masse LC, Fuemmeler BF, Anderson CB, et al. Accelerometer data reduction: a comparison of four reduction algorithms on select outcome variables. Med Sci Sports Exerc. 2005;37(11 suppl):S544-54.
13. Matthews CE, Ainsworth BE, Thompson RW, Bassett DR Jr. Sources of variance in daily physical activity levels as measured by an accelerometer. Med Sci Sports Exerc. 2002;34(8):1376-81.
14. Munneke M, de Jong Z, Zwinderman AH, Tijhuis GJ, Hazes JM, Vliet Vlieland TP. The value of a continuous ambulatory activity monitor to quantify the amount and intensity of daily activity in patients with rheumatoid arthritis. J Rheumatol. 2001;28:745-50.
15. Murphy D. Lycra working splint for the rheumatoid arthritic hand with MCP ulnar deviation. Aust J Rural Health. 1996;4(4):217-20.
16. Murphy SL, Smith DM, Clauw DJ, Alexander NB. The impact of momentary pain and fatigue on physical activity in women with osteoarthritis. Arthritis Rheum. 2008;59(6):849-56.
17. Myllykangas-Luosujarvi RA, Aho K, Isomaki HA. Mortality in rheumatoid arthritis. Semin Arthritis Rheum. 1995;25(3):193-202.
18. National Cancer Institute. Risk Factor Monitoring and Methods: SAS Programs for Analyzing NHANES 2003-2004 Accelerometer Data [Internet]. [cited 2008 Dec 3]. Available from:
19. Pincus T. The underestimated long term medical and economic consequences of rheumatoid arthritis. Drugs. 1995;50(suppl 1):1-14.
20. Silman AJ. Epidemiology of rheumatoid arthritis. APMIS. 1994;102(10):721-8.
21. Swartz AM, Strath SJ, Bassett DR Jr, O'Brien WL, King GA, Ainsworth BE. Estimation of energy expenditure using CSA accelerometers at hip and wrist sites. Med Sci Sports Exerc. 2000;32(9 suppl):S450-6.
22. Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008;40(1):181-8.
23. Welk GJ, Schaben JA, Morrow JR Jr. Reliability of accelerometry-based activity monitors: a generalizability study. Med Sci Sports Exerc. 2004;36(9):1637-45.


©2010The American College of Sports Medicine