Accelerometer-Based Physical Activity Patterns and Associations with Outcomes among Individuals with Osteoarthritis

Abstract

Background:

This study examined patterns of physical activity and associations with pain, function, fatigue as well as sleep disturbance among individuals with knee or hip osteoarthritis (OA).

Methods:

Participants (n=54) were enrolled in a telephone-based physical activity coaching intervention trial; all data were collected at baseline. Self-reported measures of pain and function (WOMAC subscales), fatigue (10-point numeric rating scale), and PROMIS sleep disturbance were collected via telephone. Accelerometers were mailed to participants and were worn for at least 3 days. Proportion of time participants spent in sedentary behavior during the morning (from wake until 12:00 pm), afternoon (12:00 pm until 5:59 pm), and evening (6:00 pm until sleep) each day was averaged across all days of wear. Pearson correlations assessed associations between activity and self-reported measures.

Results:

Participants spent a large proportion of time in sedentary behavior: 65.6% of mornings, 70.0% of afternoons, and 76.6% of evenings. Associations between proportion of time spent in sedentary behavior and reported outcomes were generally strongest in the afternoon, strongest for WOMAC function, and lowest for PROMIS sleep disturbance. In the evening hours, sedentary time was most strongly associated with fatigue.

Conclusions:

Overall, findings stress the importance of reducing sedentary behavior among adults with OA and suggest behavioral interventions may be strengthened by considering patients’ within-day variation in symptoms and activity.

BACKGROUND

Osteoarthritis (OA) is the most common joint disorder, affecting over 32.5 million adults in the US and rising in prevalence ^1,2. Individuals with OA often struggle with daily pain, activity avoidance, lack of sleep, and overall decreased quality of life ³. Physical activity (PA) is a key component of improving function, pain, and other outcomes such as mood and sleep quality in the context of knee OA. The US Department of Health and Human Services and other organizations recommend 150 minutes of moderate to vigorous physical activity (MVPA) weekly⁵, but this can be difficult for many individuals with OA to obtain ^6,7. There is some evidence that PA levels below general recommendations can still benefit individuals with OA, including reduced risk for functional limitations ^8,9. However, we still know relatively little about how specific PA patterns are associated with outcomes like pain, sleep, and fatigue among individuals with OA.

Most PA studies have focused on total daily minutes spent in different behaviors, including MVPA and sedentary behavior ^7,9,10. However, it is well recognized that PA patterns vary throughout the day ¹¹, and some studies have evaluated PA behavior specifically during morning, afternoon, and evening periods ^12–14. This is of interest because it can help to identify typical periods of sedentary behavior, which may be specific targets for intervention. In addition, some studies have shown that the relationship of PA with patient- and environmental-level characteristics varies by time of day^13–15. From clinical and public health perspectives, this information is useful in tailoring interventions and forming specific recommendations for patients. However, little is known about PA patterns across the day among individuals with OA or other chronic pain conditions^12,15,16.

In this study, we sought to describe PA patterns among individuals with hip and knee OA during morning, afternoon, and evening time periods. In addition, we were interested in the associations of these patterns with OA-related outcomes including pain, function, sleep, and fatigue, which have been associated with PA in previous studies ^15,16. However, to our knowledge prior studies have not examined associations of these variables specifically with PA during morning, afternoon, and evening periods.

METHODS

Study Design

This was a secondary analysis of baseline data from a single-group, pre-post design pilot study; study methods and primary results have been published previously ¹⁹. Briefly, the study intervention involved three PA coaching calls (focused on goal setting), three check-in emails, and linkage with community-based or online resources to support PA. The study was approved by the Institutional Review Board of the University of North Carolina at Chapel Hill, the trial was registered at clinicaltrials.gov (NCT03780400) prior to study enrollment.

Study Sample

The OA-PCP program was developed to be compatible with delivery in the context of the Center for Medicare and Medicaid Services Chronic Care Management services, with the rationale that chronic care managers could deliver the content of OA-PCP within broader Chronic Care Management phone calls. Therefore, study eligibility criteria were designed with this in mind. Participants had to be age 65 or older, and have, in addition to a diagnosis of hip and/or knee OA, at least one other chronic health condition that qualified under Chronic Care Management guidelines. Participants had to report having current symptoms in a joint with OA, using the following validated item: “Do you have pain, aching or stiffness in your knees/hips on most days?” ²⁰. Participants also had to self-report a pain score of ≥3 on a 0–10 numeric scale (0=no pain, 10=extreme pain) ²¹. Because the intervention was focused on increasing physical activity level, we only included individuals who self-reported not currently meeting the Department of Health and Human Services recommendation of at least 150 minutes MVPA per week ⁵. We used the PA Vital Sign measure to screen potential participants for PA level ^22,23. Exclusion criteria, previously reported, were based on safety issues related to engaging in an independent exercise program, as well as health conditions that would make it infeasible to complete other aspects of the study ¹⁹.

Potential participants were first identified from among patients of participating primary care providers in three University of North Carolina clinics, using the electronic medical record. We identified patients age 65 and older with diagnosis codes for knee, hip or generalized OA, a diagnosis code for at least one qualifying comorbid health condition under Chronic Care Management guidelines, and no diagnosis codes for exclusionary health conditions. Because we included the diagnosis code for generalized OA (to avoid missing a large number of patients with knee or hip OA), we also conducted a chart review to verify the presence of OA in a knee or hip. Primary care providers reviewed lists of patients eligible based on the electronic medical record and approved a final list of patients to contact. We mailed these patients an introductory letter, signed by their primary care provider, and then a study team member called patients to further assess eligibility. Patients who were eligible and interested in participating completed a verbal consent process and were mailed a Health Insurance Portability and Accountability Act waiver form to sign and return. Then participants completed baseline assessments via telephone and were mailed an accelerometer for PA assessment. Participants were paid $25 for completing each phone-based assessment and $15 for returning the accelerometer at each time point. Following return of the accelerometer at baseline, participants began the OA-PCP intervention, although post OA-PCP data are not used in this analysis.

Measures

Physical Activity Assessment

Prior to receiving the OA-PCP intervention, each participant was asked to wear an Actigraph GT3X+ (Pensacola, FL) on an elastic belt, or belt clip, during waking hours for 7 days. Instructions were provided over the phone and mailed with the accelerometer in a pre-stamped / addressed return envelope. Upon return, accelerometer data were downloaded, summarized to 60-second epochs and processed to identify wear, non-wear, and wake periods using current algorithms, participant logs, and visual inspection of data ^24,25. Minutes classified as non-wear were removed before estimating total daily waking wear and intensity outcomes. A valid day needed 8+ hours of waking wear, and participants with 3+ valid days were included in the analysis. Most participants (96%) had 5 or more days of wear. Data for the 3 participants with 3 or 4 days were examined and determined to be complete, consistent, and acceptable. Participants averaged 13.7 (standard deviation=2.3) hours of wear per day. Total minutes spent in sedentary behavior (< 100 counts per minute), light intensity PA (100–2019 counts per minute), and MVPA (2020+ counts per minute) were determined using common cut-points ^25,26.

Minutes spent at each activity level were calculated overall and for three distinct time periods for each valid day of measurement: morning (from wake until 12:00 pm), afternoon (from 12:00 pm until 5:59 pm), and evening (from 6:00 pm until sleep). To account for differences in wear time across participants during each of these time periods, we converted raw minutes to the proportion of time spent in each behavior (e.g., sedentary, light intensity PA, and MVPA). In order to determine if day of wear should be considered in the analysis, participants’ patterns of time spent in sedentary behavior, light PA, and MVPA were examined across days. We observed a high degree of within-person consistency of time spent in PA and sedentary behavior over both days of observation (e.g. day1 vs day6) and day of week (e.g. weekday vs weekend). Therefore, we averaged minutes spent in each behavior across days for each participant. Due to the very low amount and variability of MVPA in this sample (less than 5 minutes per day), sedentary time was selected as the primary outcome for analysis.

Self-Report Measures

Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC):

We assessed self-reported pain and function using WOMAC subscales ²⁷. The pain subscale includes 5 items, and the function subscale includes 17 items, all rated on a Likert scale of 0 (no pain or difficulty) to 4 (extreme pain or difficulty). Higher scores indicate worse pain and function. The reliability and validity of the WOMAC total score and subscales have been confirmed ²⁹. In patients with hip or knee OA, Bellamy et al. reported internal consistency coefficient (Cronbach’s alpha) between 0.86 and 0.95 on WOMAC pain, stiffness, and function subscales. Construct validity has been confirmed by a significant association with the Lequesne Algofunctional Index for Knees ²⁹.

PROMIS Sleep Disturbance Short Form 4a:

This scale assesses sleep quality, sleep depth, and restoration associated with sleep ³⁰. It includes 4 items, with 5 response options ranging in value from 1 to 5. Results are presented as T-scores, with higher values indicating greater sleep disturbance. This scale was shown high correlation with the full PROMIS Sleep Disturbance bank score, high reliability, and construct validity related to general health and quality of life ³¹.

Stanford Visual Numeric Rating Scale (NRS) for Fatigue:

This is a single item measure, rated on a Likert scale of 0 (no fatigue) to 10 (severe fatigue) in the past 2 weeks ²⁹. This measure is based on a pain visual NRS, which has strong correlations with health distress and general health ³³.

Participant Characteristics:

We collected the following self-reported information to characterize the study sample: age, race / ethnicity, gender, education, work status, marital status, body mass index, comorbidities³⁰, number of joints with arthritis symptoms, and duration of knee / hip OA symptoms.

Statistical Analyses

Descriptive statistics were calculated for participant characteristics using means and standard deviations for continuous variables and frequencies and percentages for categorical variables. Pearson’s correlation coefficients were calculated to determine the strength of associations between proportion of sedentary time (during morning, afternoon, and evening intervals) and each self-report measure (WOMAC pain, WOMAC function, PROMIS sleep disturbance, and fatigue NRS). As an exploratory analysis, we examined period of the day (morning, afternoon, evening) as a repeated effect to see if significant interactions existed between OA outcomes and time of day. To illustrate the relationship of WOMAC pain, WOMAC function, and fatigue scores with sedentary behaviors, we present box plots showing the proportion of sedentary time by quartile of each OA outcome.

RESULTS

Participants and Recruitment

We previously reported details regarding recruitment and enrollment into this study, and these data are summarized in Figure 1. Briefly, we identified 823 potentially eligible patients in the University of North Carolina Healthcare System electronic medical records and mailed introductory letters to 442 of these individuals who were eligible based on chart review. Among these patients, 70 screened eligible and 60 consented and completed baseline assessments. An additional 6 participants were excluded from these analyses because they did not have adequate accelerometer data based on metrics described above; there were no substantial differences in demographic or clinical characteristics between participants with and without adequate accelerometer data (data not shown). Of the remaining 54 participants in these analyses, 80% were women, 69% were White, the mean age was 72.8 ± 5.7 years old, and the mean BMI was 33.1 ± 7.2 kg/m² (Table 2).

Figure 1.

OA-PCP Study Participant Flow

Table 2.

Average times spent in each activity type throughout the day.

	Morning			Afternoon			Evening
Activity Type	Avg. Activity Mins (SD)	Avg. Mins of Wear (SD)	% of time activity in wear (range)	Avg. Activity Mins (SD)	Avg. Mins of Wear (SD)	% of time activity in wear (range)	Avg. Activity Mins (SD)	Avg. Mins of Wear (SD)	% of time activity in wear (range)
Sedentary	145.7 (55.3)	222.2 (71.0)	65.6 (35.2–91.6)	243.6 (34.0)	348.2 (17.4)	70.0 (54.4–90.4)	201.7 (65.6)	262.0 (74.4)	76.6 (52.7–90.8)
Light Activity	73.2 (32.8)	222.2 (71.0)	33.0 (8.4–54.4)	101.5 (33.5)	348.2 (17.4)	29.1 (9.6–45.6)	59.3 (24.9)	262.0 (74.4)	23.0 (9.2–46.0)
MVPA	3.3 (5.7)	222.2 (71.0)	1.4 (0–10.4)	3.0 (4.8)	348.2 (17.4)	0.9 (0–6.2)	1.0 (1.3)	262.0 (74.4)	0.4 (0–1.9)

Physical Activity Patterns

Table 2 shows minutes spent in each PA category, accelerometer wear time, and percent time spent in each PA category for morning, afternoon, and evening hours. Low levels of MVPA were noted throughout the day (1.4% of mornings, 0.9% of afternoons, and 0.4% of evenings). Most of the day was spent in sedentary behavior, with some variation across times of day. Participants spent less time engaging in sedentary behavior in the morning with a gradual increase as the day continued (65.6% of mornings, 70% of afternoons, and 76.6% of evenings).

Associations of Physical Activity Patterns with Self-Report Measures

Table 3 shows the correlations between proportion of time spent in sedentary behavior and self-report measures for morning, afternoon, and evening; overall these correlations were small to moderate. Correlations with proportion of sedentary time were highest for the WOMAC function subscale (r=0.11–0.40). On the other hand, PROMIS Sleep Disturbance scores were the least correlated with sedentary behavior throughout the day (r= −0.10–0.11). Overall, sedentary behavior in the afternoon was more strongly correlated with self-report measures compared to morning or evening sedentary time; evening activity patterns had the lowest correlations with self-report measures overall. In repeated measures analyses (results not shown), sedentary behavior was not associated with the WOMAC pain subscale, PROMIS Sleep Disturbance scores, or fatigue, nor were interactions with the time period. A significant time-by-WOMAC function interaction term (p=0.004) indicated that the effect of WOMAC function differed across time periods, which supports our findings in simpler analyses. However, these results are exploratory because the study is not sufficiently powered to test for an interaction.

Table 3.

Associations between proportion of sedentary time and patient outcomes.

	Morning		Afternoon		Evening
Patient Outcome	Pearson Correlation	p-Value	Pearson Correlation	p-Value	Pearson Correlation	p-Value
WOMAC Function	0.30	0.03	0.40	0.003	0.11	0.42
WOMAC Pain	0.18	0.19	0.30	0.02	0.05	0.73
Fatigue	0.26	0.06	0.30	0.03	0.21	0.13
*PROMIS Sleep Disturbance	−0.10	0.49	0.04	0.79	0.11	0.42

^*PROMIS Sleep Disturbance score missing for 1 participant (n=53)

Figure 2 illustrates the proportion of afternoon time spent in sedentary behavior, with participants grouped by quartiles of WOMAC function score, WOMAC pain score, and fatigue score. While it is difficult to identify a strong linear pattern in our small sample, the proportion of sedentary behavior tended to increase in participants having the poorest outcomes. These results should be viewed with caution due to the small numbers of participants in the second quartiles (n = 8 for WOMAC pain, n = 12 for WOMAC Function, and n = 5 for Fatigue) and high variation of afternoon PA in all groups. There was some consideration to illustrate these results as tertiles rather than quartiles due to the limited data present across the second quartiles, however, we felt it was more important to include the second quartiles in order to fully represent our sample.

Figure 2.

Proportion of afternoon time spent in sedentary behavior, based on quartiles of (A) WOMAC pain scores, (B) WOMAC function scores, and (C) fatigue scores. The boxplots show the mean (◇) and median (horizontal line drawn within the box); the dots are individual data points. The “whiskers” (the lines extending parallel from the boxes) indicate variability outside the upper and lower quartiles.

DISCUSSION

In this study, we assessed patterns of PA and sedentary behavior, as well as their associations with patient outcomes among individuals with knee or hip OA, with particular attention to patterns across different times of day. The first main finding was that participants spent the majority of their time in sedentary behavior (71% overall), with very little MVPA (mean 7.4 minutes/ day). These findings are consistent with other studies of PA patterns among individuals with lower extremity OA ^32–34. Among individuals with or at risk for knee OA in the Osteoarthritis Initiative, 66% of waking time was spent in sedentary behavior, and the average amount of daily moderate or vigorous activity was less than 20 minutes ³². Activity levels were somewhat lower among participants in our study, and this may be due to a higher symptom burden compared with Osteoarthritis Initiative, some of whom did not have symptomatic OA. In a small study of German patients with knee OA, sedentary time was also high, with 88% of time spent in non-locomotion behavior ³⁴. Results of these studies, along with ours, indicate the urgency of efforts to increase activity and reduce sedentary time among individuals with knee OA. This is particularly important given the associations of sedentary behavior on future functional limitations and frailty in these patients^9,32.

Participants in our study were somewhat more active in the morning, with the greatest average proportion of sedentary behavior occurring in the evening; however, differences in the proportion of sedentary time across times of day were modest. Few studies have examined PA patterns across times of day among individuals with OA. However, findings of our study are consistent with prior research in this area. For example, in a sample of individuals with hip or knee OA, average minutes of activity was highest from waking time until 11:00 am, decreasing as the day progressed, with lower activity levels from 7:00 pm until bedtime ¹⁵; similar patterns were observed in a sample of women with hip or knee OA ³⁴. In studies of older adults, PA was also highest in the morning, with increasing sedentary behavior throughout the day ^36.37. Findings of these studies suggest that for older individuals and those with chronic pain, it may be easier to be active in the morning, as pain and/or fatigue may worsen as the day continues ³⁵. While relationships of pain- and fatigue-related activity interference with PA behaviors are complex and may vary throughout the day, results of our study and others suggest it may be useful to address time of day in behavioral interventions aimed at increasing PA among individuals with OA. Specifically, PA counselors can help patients to consider times of day when pain and fatigue are lowest, as these may be the best times to engage in purposeful activity. However, patients should also be encouraged to consider feasible strategies for decreasing sedentary behavior during times when this tends to be most common (e.g., evening), given the detrimental impacts of high levels of sedentary behavior.

Correlations of patient outcomes with sedentary behavior in this study were relatively small. This is consistent with a prior study finding relatively small associations between pain and fatigue with average weekly PA ¹⁵. However, we observed somewhat stronger associations of function with sedentary time, particularly in the morning and afternoon. Because this was a cross-sectional study, the direction of this association cannot be determined. It is plausible that lower function led to greater sedentary behavior or the converse, and it is likely a bi-directional relationship ³⁸. It is also interesting that associations of patient outcomes with sedentary behavior tended to be strongest in the afternoon. Murphy et al. found that among individuals with knee OA, differences in activity between those with low and high pain tended to be more accentuated during the afternoon hours, particularly relative to evening hours ¹⁵. Again, it is not possible to ascertain causality from our cross-sectional data; however, it is possible that symptoms are more likely to influence PA patterns during the afternoon time period compared with other times of day. During evening hours, fatigue was most strongly associated with the proportion of time spent in sedentary behavior. Other research has also found fatigue to be an important predictor of physical activity among individuals with OA ^17,35. Our study’s finding that fatigue may be particularly relevant in the evening hours has implications for PA counseling. In particular, fatigue may be a particularly important barrier to address with respect to reducing sedentary behavior in the evening, when it is most common. The small correlation of sleep disturbance with PA was surprising, giving prior research showing that poor sleep quality was associated with less activity among individuals with or at risk for OA ¹⁶. This may be due to different measures of sleep quality or the overall lower level of PA in our study sample.

There are several limitations to this study. This was a small sample of 54 patients, and individuals who consented to participate in a trial of PA intervention may differ in their activity patterns relative to the general population of patients with OA. In addition, many patients who were contacted chose not to participate in the study; although this is common in clinical trials, there may have been differences between participants and non-participants with respect to baseline PA patterns. Along with the small sample size, the majority of the sample were female, Caucasian, and had some college education, which may also limit generalizability. Because wear time varied, there are some differences in the total number of minutes each person contributed to the 3 time periods examined. However, participants contributed substantial data to each time period, averaging 3.5 to 5.8 hours of wear. In addition, our approach of examining the proportion of time (rather than raw minutes) helps to account for differences in wear. Finally, other studies of PA in patients with knee OA have identified predictors that were not measured in our study, such as coping, self-efficacy, and mood ^39,40.

CONCLUSION

In summary, we observed high levels of sedentary behavior among patients with knee or hip OA, which was somewhat accentuated during the evening hours, and small to modest associations of patient outcomes with activity, with strongest associations for physical function. This study is one of few to explore PA and sedentary behavior patterns across times of day among patients with OA, and further research in this area could help to inform behavioral interventions ^12,41. Continued and more complex analyses of how outcomes such as pain, function, and fatigue influence PA patterns at different times of day could be particularly fruitful. Overall, our findings stress the importance of continuing efforts to enhance PA among individuals with knee OA and suggest that patient outcomes (e.g., pain, function, and fatigue), as well as daily variations in activity, should be considered within behavioral interventions.

Table 1.

Baseline Participant Characteristics (N=54)

Mean age (SD), years	72.8 (5.7)
Gender
N(%) Female	43 (80%)
Race
N(%) White	37 (69%)
N(%) Black / African American	14 (26%)
N(%) Asian	1 (2%)
N(%) Other	1 (2%)
Ethnicity
N(%) Hispanic	2 (4%)
Education
N(%) with at least some college	44 (82%)
Marital Status
N(%) Married or living with partner	35 (65%)
Work Status
N(%) Working full or part time	16 (30%)
Mean (SD) Body mass index, kg/m2	33.1 (7.2)
Mean (SD) Number of comorbid illnesses*	4.2 (1.7)
Mean (SD) Years with arthritis symptoms	11.1 (9.6)
Mean (SD) Number joints with arthritis symptoms	7.8 (2.8)
Mean (SD) WOMAC Function Score	26.3 (13.3)
Mean (SD) WOMAC Pain Score	8.0 (3.7)
Mean Fatigue Score	5.2 (2.4)
Mean PROMIS Sleep Disturbance Score	53.1 (2.6)

^*Possible range: 0–16. Missing data: Race: n=1; Ethnicity: n=8; PROMIS Sleep Disturbance n=1

ABBREVIATIONS

MVPA

Moderate to Vigorous Physical Activity

NRS

Numeric Rating Scale

OA

Osteoarthritis

OA-PCP

Osteoarthritis Physical Activity Care Pathway

PA

Physical Activity

PROMIS

Patient Reported Outcomes Measurement Information System

WOMAC

Western Ontario and McMaster Universities Osteoarthritis Index

Availability of data and material:

The full data sets generated during the current study are not publicly available because they include personally identifiable data from study participants. However, de-identified data are available from the corresponding author on reasonable request.