Ecological momentary pain and physical activity volume and patterns in older adults with chronic multisite pain

Abstract

Background

Chronic multisite musculoskeletal pain is prevalent in older adults and contributes to disability, partly through restriction and dysregulation of physical activity (PA). Prior studies have relied primarily on static pain measures and summary PA metrics, leaving time-varying pain-activity relationships poorly understood. This observational PRIME pilot study examined associations of baseline and time-varying pain characteristics with PA volume, fragmentation, and diurnal activity patterns in older adults with chronic multisite pain.

Methods

Thirty community-dwelling adults aged 65 years or older completed 14 days of wrist-worn accelerometry and ecological momentary assessment (EMA) of pain four times daily. PA outcomes included total activity counts/day (TAC), activity fragmentation, active minutes/day, and diurnal patterns. Linear regression and mixed-effects models examined associations of pain characteristics with PA, adjusting for age and sex.

Results

EMA adherence was high (96% prompt completion). Back pain at baseline was associated with lower TAC (β = − 388,910, p = 0.029), fewer active minutes (β = − 74.5, p = 0.016), greater fragmentation (β = 5.38%, p < 0.001), and reduced afternoon/evening activity. Higher momentary pain severity was associated with lower activity from wake-up to the morning prompt (β = − 15,688, p = 0.045) and a positive trend during the overnight interval (β = 12,169, p = 0.087). Mixed-effects models showed significant three-way interactions among overall pain severity, time (linear and quadratic), and weekday/weekend status (p < 0.04), indicating contextual modulation of diurnal PA trajectories.

Conclusions

Pain-activity relationships in older adults are heterogeneous and context-dependent. Integrating EMA with accelerometry may inform personalized interventions to maintain mobility and physical functioning.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12877-026-07971-0.

Introduction

Chronic pain affects more than half of community-dwelling adults aged 65 years and older [1] and represents a major contributor to disability, reduced quality of life, and increased healthcare utilization in aging populations. Musculoskeletal pain, in particular, has been consistently associated with impaired physical functioning, mobility limitation, falls, depression, and cognitive decline in later life [2–8]. A key pathway through which chronic pain may contribute to these adverse outcomes is restriction and dysregulation of daily physical activity (PA), as older adults may intentionally or subconsciously limit movement to avoid pain exacerbation.

Current research on pain and PA in older adults has focused on pain presence or severity and has relied heavily on self-reported PA measures or accelerometer-derived summary metrics, such as step counts or time spent in moderate-to-vigorous physical activity (MVPA) [9–11]. These approaches, while informative, may fail to capture the complex and dynamic nature of daily activity in free-living environments, particularly among older adults whose activity is dominated by light-intensity movements and intermittent bouts of activity. Moreover, pain itself is not a static exposure. Pain severity, location, and laterality can fluctuate substantially within a day and across days [12–14], which may contribute to heterogeneous activity profiles that are poorly characterized by traditional cross-sectional or retrospective assessments.

Growing evidence indicates that chronic multisite pain, defined as pain occurring in two or more locations, is more strongly associated with mobility limitation, functional decline, and falls than single-site pain [3–5, 15–17] and has been proposed as a geriatric syndrome [18, 19]. Despite its high prevalence and clinical significance, multisite pain remains understudied in relation to objectively measured PA patterns. In addition, pain characteristics beyond severity, such as pain distribution and laterality, have received limited attention [20, 21]. Preliminary findings from observational cohorts suggest that unilateral lower extremity pain may be associated with more pronounced reductions in activity than bilateral pain, potentially due to asymmetry-related gait adaptations, muscle weakness, or fear of movement [22]. These observations highlight the need to move beyond global pain ratings and examine how specific pain characteristics shape daily activity behavior.

Ecological momentary assessment (EMA) offers a powerful approach to characterizing pain as a time-varying and context-dependent experience [23]. By capturing repeated and real-time reports of pain, EMA minimizes recall bias and allows for the quantification of within-day and day-to-day variability in pain characteristics [23]. Prior EMA studies in chronic pain populations have largely focused on osteoarthritis or chronic low back pain and have relied on pain severity as the main characteristic [23, 24]. Little is known about the feasibility of using EMA to capture multisite pain characteristics, including pain distribution and laterality, in older adults or how fluctuations in these characteristics relate to objectively measured PA patterns.

The Pain and Activity Rhythms In Multisite-pain Elders (PRIME) pilot study was designed to address these gaps by integrating EMA of multiple pain characteristics with accelerometer data in community-dwelling older adults with chronic multisite pain. By examining pain fluctuations and variability alongside PA volume, diurnal patterns, and activity fragmentation over a 14-day period, this study aims to provide a more comprehensive understanding of the dynamic interplay between pain and daily activity in later life.

Methods

Study design

The PRIME study is an observational pilot study designed to evaluate the feasibility and acceptability of a 14-day EMA protocol for assessing multisite pain characteristics in older adults and to explore associations between time-varying pain characteristics and objectively measured PA quantities and patterns. The study integrates repeated real-time pain assessments with continuous wrist-worn accelerometry in a free-living setting. The study was conducted between January 2024 and December 2024 and was approved by the Institutional Review Board of the University of Pittsburgh.

Study participants

A total of 30 community-dwelling adults aged 65 years and older were recruited from the Pittsburgh Claude D. Pepper Older Americans Independence Center Research Registry and the Pitt + Me research participation portal. Individuals who were interested in the study completed an initial telephone screening to assess eligibility, followed by an in-person clinical visit to confirm eligibility on cognitive function and complete baseline assessments.

Inclusion criteria were: (1) age 65 years or older; (2) presence of chronic pain at two or more musculoskeletal sites, including shoulder, elbow, hand/wrist/finger, back, hip, knee, or ankle/foot/toe; (3) ability to understand and communicate in English; (4) able to walk a quarter of mile and climb a flight of stairs without personal assistance (walk aids permitted); and (5) willingness to complete 14 days of EMA and activity monitoring. Exclusion criteria included terminal illness, acute medical conditions requiring urgent care, severe vision or hearing impairment, and cognitive impairment defined as a Montreal Cognitive Assessment (MoCA) < 23 [25].

Study procedures

At the baseline clinical visit, participants completed a cognitive function assessment using MoCA to confirm eligibility. The MoCA was designed as a rapid screening instrument for mild cognitive dysfunction [26]. It assesses different cognitive domains: attention and concentration, executive function, memory, language, visuoconstructional skills, conceptual thinking, calculations, and orientation. The total possible score is 30 points and a cutoff < 23 was used to determine cognitive impairment [25]. If participants were confirmed without cognitive impairment, a comprehensive health interview and physical assessment were administered. At the end of the visit, participants received training on EMA survey completion and were fitted with a wrist-worn accelerometer. They were instructed to begin EMA data collection the day following the clinical visit and to wear the accelerometer continuously for 14 consecutive days.

Baseline health interview and assessments

Sociodemographic information collected included age, sex, race/ethnicity, and years of education. Height and weight were assessed using a stadiometer and a calibrated scale, respectively. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Participants reported physician-diagnosed chronic conditions (e.g., cardiovascular disease, hypertension, diabetes, osteoarthritis, neuropathy, etc.). Pain severity and pain interference were assessed using the Brief Pain Inventory (BPI)–Short Form [27]. Participants were also asked to select where they have pain from the following sites: shoulder, elbow, hand/wrist/finger, back, hip, knee, or ankle/foot/toe, followed by questions about severity (scale 0–10) and laterality (left side, right side, or both) of the pain [22]. Use of all prescription and over-the-counter medications taken in the previous 2 weeks was recorded. Analgesic medications included opioid and nonopioid analgesics and daily use was determined from dose and frequency information. Depressive symptoms were measured using the Center for Epidemiologic Studies Depression Scale–Revised (CESD-R) [28]. Fear of falling was measured using the Falls Efficacy Scale–International [29]. Usual gait speed was assessed as part of the Short Physical Performance Battery (SPPB) [30].

Ecological momentary assessment of pain

Participants completed EMA surveys delivered via Qualtrics on their preferred personal devices (e.g., cellphone, tablets, or computer) for 14 days. Surveys were prompted four times per day: at self-reported wake-up time, at bedtime, and at two random times during the morning (9:00–12:00) and afternoon (13:00–18:00). Each survey assessed current overall pain intensity using a 0–10 numeric rating scale. When pain was reported (> 0), participants identified pain location(s), rated site-specific pain severity, and reported pain laterality (left, right, or bilateral) (Supplementary Material 1). If participants missed the survey, a text reminder was sent to participants in 15 min except for the bedtime point. If participants missed surveys across two days, they were contacted immediately by study staff members to remedy any potential barriers to consistent completion.

Physical activity assessment

PA was measured using a triaxial wrist-worn Ametris CentrePoint Insight Watch (CPIW) (Ametris, LLC) worn on the non-dominant wrist for 14 days, 24 h per day. The CPIW is water resistant (IPX7, 1 m); participants were instructed to wear the device continuously, including during routine water exposure, but to avoid prolonged swimming or other activities inconsistent with study instructions. Raw acceleration data were collected at 30 Hz and processed into 1-min epochs using the CenterPoint platform. Participants completed a sleep diary to aid in distinguishing sleep from wake periods. A valid day was defined as having no more than 10% missing data [31]. Participants with at least three valid days were included in analyses [22]. For valid days, missing minute-level values were imputed using the participant-specific average for the corresponding minute across other valid days [22]. PA metrics were then summarized to (1) total activity counts (TAC), (2) activity fragmentation, quantified as active-to-sedentary transition probability (ASTP), and (3) total active minutes. TAC was calculated as the sum of minute-level activity counts and averaged across all valid days. For ASTP and active minutes, each minute was classified as active if activity counts were ≥ 1,853 counts per minute [32]. Consecutive active minutes were aggregated into active bouts. ASTP was calculated as the reciprocal of the mean duration of active bouts for each participant, with higher values indicating more fragmented activity patterns and a greater likelihood of transitioning from activity to sedentary behavior [33]. Total daily active time was defined as the sum of active minutes and averaged across valid days [34].

Follow-up survey on EMA

Following completion of the 14-day monitoring period, participants completed a brief follow-up survey assessing ease of system use, clarity of survey questions, helpfulness of email and text prompts, and perceived burden of completing daily pain surveys. Likert-scale items ranged from 1 (strongly agree) to 5 (strongly disagree). Survey burden was rated on a 0–10 numeric scale, with higher values indicating greater burden (Supplementary Material 2). Responses were summarized using descriptive statistics.

Statistical analysis

Descriptive statistics were used to characterize the study sample and summarize EMA completion rates, follow-up surveys, pain characteristics, and PA measures. Linear regression models were used to examine the association between baseline pain characteristics (e.g., pain severity, pain distribution, and pain laterality) and the average PA metrics over 14 days. Given the small sample size, we prioritized parsimonious model specifications. The primary model was the unadjusted linear regression model, with sensitivity analyses adjusting for a limited set of covariates (e.g., age and sex) to avoid overfitting. Heteroskedasticity-robust (Huber-White) standard errors were reported.

Multiple linear mixed effects models tested differences in total activity counts (TAC) across four 6-h time intervals of the day (i.e., 12:00 am to 5:59 am, 6:00 am to 11:59 am, 12:00 pm to 5:59 pm, 6:00 pm to 11:59 pm) between older adults with and without pain at each site, adjusting for covariates. A compound symmetry matrix was used to account for within-participant clustering of time intervals and restricted maximum likelihood (REML) estimation was applied to fit the models.

Multilevel modeling was used to quantify within-day and day-to-day variability in pain characteristics and to examine associations between time-varying pain measures and PA outcomes across the 14-day period. We aggregated the 14 days of data by day of the week. In the mixed effects models, time of day was coded as 1, 2, 3, and 4 corresponding to the four EMA prompt time points(wake-up, morning, afternoon, and bedtime). We included both linear and quadratic terms (time*time) for time because activity counts followed a nonlinear diurnal curve, and adding the quadratic term improved model fit. The linear time term represents the early-day direction and rate of change in PA across time points, whereas the quadratic time term represents curvature in the diurnal PA trajectory. We further categorized days as weekday versus weekend and added three-way interactions among pain, time of day, and weekday/weekend status to test whether pain-activity associations differed by daily context.

All significance tests were conducted using two-sided tests. The significance level α was set as 0.05. All statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC) and R (version 4.5.1).

Results

Sample characteristics and baseline pain

Of the 49 older adults screened, 40 were eligible at the initial phone screening. Of these, 7 declined participation, 2 had scheduling conflicts, and 1 was later deemed ineligible due to cognitive impairment based on the MoCA administered at the clinical visit, yielding a final sample of 30 participants. Those who were not eligible were primarily excluded due to not having multisite pain across the following locations: shoulder, elbow, hand/wrist/finger, back, hip, knee, or ankle/foot/toe. The average age was 74.1 (standard deviation [SD] = 6.0) years. Ten (33.3%) participants were women and all participants were White. The average years of education were 18.5 (SD = 2.7). The average BPI score is 3.5 (SD = 1.7) for pain severity and 2.7 (SD = 1.7) for pain interference, respectively. Ten participants had moderate to severe pain with a BPI severity ≥ 4. The sample characteristics for all participants and by pain severity are presented in Table 1.

Table 1.

Sample characteristics by overall pain severity at baseline

	All (N = 30)	No pain or mild pain (n = 20)	Moderate to severe pain (n = 10)	p-value*
Characteristics	Mean ± SD or n (%)
Age (years)	74.1 ± 6.0	74.9 ± 6.8	72.5 ± 3.7	0.467
Female	10 (33.3)	6 (30.0)	4 (40.0)	0.690
White	30 (100.0)	20 (100.0)	10 (100.0)	1.000
Education years	18.5 ± 2.7	18.4 ± 2.7	18.8 ± 2.9	0.744
BMI	26.9 ± 3.5	26.4 ± 3.6	27.9 ± 3.3	0.267
Cardiovascular disease	12 (40.0)	8 (40.0)	4 (40.0)	1.000
Hypertension	10 (33.3)	6 (30.0)	4 (40.0)	0.690
Diabetes	4 (13.3)	2 (10.0)	2 (20.0)	0.584
Osteoarthritis	23 (76.7)	16 (80.0)	7 (70.0)	0.657
Neuropathy (3 missingness)	4 (13.3)	2 (11.8)	2 (20.0)	0.613
Depression	3 (10.0)	1 (5.0)	2 (20.0)	0.251
MoCA score	26.1 ± 2.3	26.2 ± 2.2	26.5 ± 2.5	0.635
Usual gait speed (m/s)	0.9 ± 0.2	0.9 ± 0.2	0.9 ± 0.2	0.746
Fear of falling	24.6 ± 6.7	23.5 ± 4.9	26.7 ± 9.3	0.415
Low concern	8 (26.7)	5 (25.0)	3 (30.0)	0.478
Moderate concern	14 (46.7)	11 (55.0)	3 (30.0)
High concern	8 (26.7)	4 (20.0)	4 (40.0)
Daily analgesics use	10 (33.3)	6 (30.0)	4 (40.0)	0.690

Values are mean ± SD for continuous variables and n (%) for categorical variables unless otherwise indicated

SD Standard deviation, BMI Body mass index, MoCA Montreal cognitive assessment

*Mann–Whitney U test was used for continuous variables and Fisher’s Exact test was used for categorical variables

For baseline pain characteristics, the majority (66.7%) of participants had wrist/hand/finger pain (Fig. 1), followed by back, knee, hip, shoulder, ankle/foot/toe, and elbow pain. Participants tended to report more severe pain in hip, followed by knee, back, and wrist/hand/finger pain. Participants with hip pain tended to have unilateral pain while those with knee pain tended to have bilateral pain (Fig. 1).

Fig. 1.

Baseline pain characteristics. A Percentage of participants reporting pain at each musculoskeletal site. B Mean pain severity by location among participants reporting pain at that site. C Pain laterality distribution by site, categorized as no pain, unilateral pain, or bilateral pain

Associations between baseline pain and averaged physical activities across 14 days

All participants had 14 valid days of accelerometer assessment. The average daily activity measures during the following 14 days were 1,972,937 for TAC, 23.4% for ASTP, and 384 for active minutes. BPI pain severity and interference were not significantly associated with average activity during the 14 days (Table 2). We found that participants with elbow pain had higher TAC (β = 547,166, p = 0.027), lower ASTP (β = −4.43, p < 0.001), and more active minutes (β = 93.76, p = 0.016) compared with those without elbow pain. Back pain was associated with 480,111 fewer activity counts, 5.34% higher ASTP, and 85.64 fewer active minutes (Table 2). Similar patterns were observed for site-specific pain severity. For pain laterality, bilateral elbow pain was associated with 790,012 more activity counts, 4.69% lower activity fragmentation, and 160.06 more active minutes per day, compared with those without elbow pain. Unilateral ankle/foot/toe pain was associated with 871,686 more activity counts and 125.69 more active minutes per day than no ankle/foot/toe pain (Table 2). Linear regression models adjusted for age and sex did not substantially change the results (Supplementary Table 1).

Table 2.

Pain characteristics at baseline and average daily physical activity over 14 days

	TAC (× 1000)			ASTP			Active time (mins)
	β	Robust SE	p-value*	β	Robust SE	p-value*	β	Robust SE	p-value*
BPI pain severity	−40.20	53.15	0.456	0.71	0.58	0.235	−7.64	9.22	0.414
BPI pain interference	41.16	64.07	0.526	0.16	0.53	0.765	6.23	10.36	0.553
Pain presence at each location
Shoulder pain	344.05	198.22	0.094	−2.93	1.81	0.116	61.90	33.51	0.075
Elbow pain	547.17	233.99	0.027	−4.43	1.06	< 0.001	93.76	36.67	0.016
Wrist/hand/finger pain	27.33	197.27	0.891	−0.06	1.71	0.971	0.58	35.24	0.987
Back pain	−480.11	192.44	0.019	5.34	1.55	0.002	−85.64	31.72	0.012
Hip pain	−40.86	204.03	0.843	1.97	1.75	0.269	−19.85	35.32	0.579
Knee pain	206.98	199.41	0.308	−1.75	1.74	0.324	47.22	33.54	0.170
Foot/ankle/toe pain	270.45	248.67	0.286	−1.50	2.17	0.497	52.87	36.82	0.162
Pain severity at each location
Shoulder pain	47.04	40.24	0.252	−0.40	0.43	0.363	8.35	7.21	0.257
Elbow pain	202.64	67.48	0.006	−1.59	0.44	0.001	35.85	11.43	0.004
Wrist/hand/finger pain	30.26	40.80	0.465	−0.13	0.38	0.730	2.22	7.24	0.761
Back pain	−74.99	26.98	0.010	1.00	0.25	0.001	−12.92	4.51	0.008
Hip pain	−16.30	30.85	0.602	0.37	0.26	0.172	−4.66	5.47	0.401
Knee pain	25.21	33.53	0.459	−0.12	0.28	0.658	5.37	5.50	0.337
Foot/ankle/toe pain	33.67	58.63	0.570	−0.11	0.54	0.841	7.44	9.02	0.417
Pain laterality at each location
Shoulder pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	340.02	211.24	0.119	−3.37	1.71	0.058	68.20	35.52	0.066
Bilateral	357.49	394.99	0.373	−1.45	4.81	0.765	40.89	61.79	0.514
Elbow pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	486.46	273.68	0.087	−4.37	1.09	< 0.001	77.18	39.60	0.062
Bilateral	790.01	103.617	< 0.001	−4.69	0.99	< 0.001	160.06	18.39	< 0.001
Wrist/hand/finger pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	−60.53	251.55	0.812	1.17	2.46	0.637	−12.26	41.83	0.772
Bilateral	85.90	224.30	0.705	−0.89	1.79	0.625	9.13	40.78	0.824
Hip pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	−74.71	213.90	0.730	2.01	1.89	0.296	−27.01	37.47	0.477
Bilateral	60.69	299.53	0.841	1.85	2.46	0.459	1.64	47.46	0.973
Knee pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	82.47	221.10	0.712	0.38	2.62	0.884	19.75	39.86	0.624
Bilateral	258.87	226.85	0.264	−2.64	1.82	0.158	58.67	38.34	0.138
Ankle/foot/toe pain
No	ref	ref	ref	ref	ref	ref	ref	ref	ref
Unilateral	871.69	238.87	0.001	−6.34	3.22	0.059	125.69	45.94	0.011
Bilateral	29.96	249.38	0.905	0.44	2.06	0.832	23.74	37.49	0.532

TAC Total activity counts, ASTP Active-to-sedentary transition probability, BPI Brief Pain Inventory

*P-values were based on heteroskedasticity-robust (Huber–White) standard errors (SEs). Conventional ordinary least squares SEs yielded similar conclusions

Bolded values indicate statistical signficant results (p<0.05)

For diurnal patterns of daily activity averaged during the 14 days, we found that participants with elbow pain had more activities during the afternoon (12:00 pm to 5:59 pm; β = 292,310, p = 0.002), while those with back pain tended to have less activity during the afternoon (β = −242,714, p < 0.001) and evening (6:00 pm to 11:59 pm; β = −136,106, p = 0.046) (Table 3). Moderate to severe back pain was associated with less activity during the morning (6:00 am to 11:59 am; β = −152,070, p = 0.029) and afternoon (β = −169,373, p = 0.015). For pain laterality, those with unilateral or bilateral elbow pain had significantly more activities during afternoon than those without elbow pain (p < 0.02). Unilateral ankle/foot/toe pain was associated with more activities during the morning (β = 400,251, p = 0.005) and evening (β = 312,071, p = 0.026) (Table 3). Differences in diurnal patterns of activities by pain characteristics are presented in Fig. 2.

Table 3.

Interaction effect of 6-h time intervals and baseline pain characteristics on total daily activity counts (TAC) for each time interval during follow-up

	12:00 am to 5:59 am	6:00 am to 11:59 am	12:00 pm to 5:59 pm	6:00 pm to 11:59 pm
	Beta Coefficient* (× 1000 counts) (p-value)
BPI pain severity
None or mild	ref	ref	ref	ref
Moderate to severe	−3.91 (0.959)	−113.80 (0.133)	−36.70 (0.626)	−22.55 (0.764)
BPI pain interference
None or mild	ref	ref	ref	ref
Moderate to severe	11.08 (0.900)	−113.16 (0.204)	−45.59 (0.607)	−83.09 (0.350)
Pain presence
Shoulder pain	−7.75 (0.912)	113.95 (0.105)	134.26 (0.057)	107.29 (0.126)
Elbow pain	−33.77 (0.709)	159.33 (0.081)	292.31 (0.002)	88.25 (0.331)
Wrist/hand/finger pain	34.76 (0.646)	19.14 (0.800)	−46.93 (0.536)	53.81 (0.478)
Back pain	−12.68 (0.851)	−117.23 (0.085)	−242.71 (< 0.001)	−136.11 (0.046)
Hip pain	36.28 (0.612)	−60.53 (0.399)	4.65 (0.948)	28.08 (0.695)
Knee pain	14.48 (0.840)	62.13 (0.386)	89.04 (0.216)	34.44 (0.631)
Ankle/foot/toe pain	−7.80 (0.926)	92.72 (0.268)	78.05 (0.351)	105.36 (0.209)
Pain severity at each location (moderate to severe vs none/mild pain)
Shoulder pain	9.51 (0.905)	97.51 (0.223)	87.37 (0.275)	94.00 (0.240)
Elbow pain	−59.73 (0.760)	203.21 (0.301)	371.04 (0.061)	48.21 (0.805)
Wrist/hand/finger pain	4.66 (0.949)	−49.20 (0.499)	19.90 (0.784)	79.50 (0.276)
Back pain	−13.05 (0.849)	−152.07 (0.029)	−169.37 (0.015)	−85.37 (0.216)
Hip pain	36.28 (0.612)	−60.53 (0.399)	4.65 (0.948)	28.08 (0.695)
Knee pain	14.41 (0.842)	49.81 (0.492)	96.97 (0.183)	47.86 (0.509)
Ankle/foot/toe pain	−3.86 (0.970)	69.04 (0.508)	100.50 (0.335)	140.61 (0.179)
Pain laterality
Shoulder pain
No	ref	ref	ref	ref
Unilateral	−16.18 (0.833)	111.92 (0.147)	128.82 (0.096)	110.09 (0.154)
Bilateral	20.37 (0.866)	120.75 (0.318)	152.36 (0.209)	97.93 (0.418)
Elbow pain
No	ref	ref	ref	ref
Unilateral	−34.20 (0.733)	182.47 (0.071)	239.04 (0.019)	55.61 (0.579)
Bilateral	−32.06 (0.866)	66.77 (0.725)	505.40 (0.009)	218.84 (0.250)
Wrist/hand/finger pain
No	ref	ref	ref	ref
Unilateral	33.21 (0.722)	31.50 (0.736)	−103.72 (0.269)	−6.00 (0.949)
Bilateral	35.80 (0.671)	10.91 (0.897)	−9.07 (0.914)	93.68 (0.268)
Hip pain
No	ref	ref	ref	ref
Unilateral	32.35 (0.678)	−53.26 (0.495)	−27.05 (0.728)	22.43 (0.773)
Bilateral	48.07 (0.668)	−82.37 (0.464)	99.75 (0.375)	45.02 (0.688)
Knee pain
No	ref	ref	ref	ref
Unilateral	4.16 (0.968)	76.03 (0.461)	23.79 (0.817)	−54.07 (0.600)
Bilateral	18.78 (0.811)	56.34 (0.473)	116.23 (0.141)	71.32 (0.364)
Ankle/foot/toe pain
No	ref	ref	ref	ref
Unilateral	−48.07 (0.727)	400.25 (0.005)	187.47 (0.175)	312.07 (0.026)
Bilateral	8.31 (0.928)	−30.29 (0.742)	34.28 (0.710)	22.68 (0.805)

BPI Brief Pain Inventory

^*Estimates of differences in activity counts by pain characteristics during each time interval

Bolded values indicate statistical signficant results (p<0.05)

Fig. 2.

Diurnal patterns of daily physical activity by baseline pain characteristics. Activity counts per minute were averaged across 14 days and plotted by time of day. A Brief Pain Inventory (BPI) pain severity and pain interference. B Pain presence at each site. C Pain severity at each site. D Pain laterality at each site

Ecological momentary assessment of pain for 14 days

For feasibility of EMA data collection, participants were scheduled to complete four EMA surveys per day for 14 days (56 prompts per participant). Overall, 1611 of 1680 expected prompts were completed, yielding an overall completion rate of 96%. At the participant level, the mean (SD) completion rate was 96% (SD = 7.3%), with 90% of participants completing at least 80% of prompts. Completion rates were highest for wake-up prompts (97.4%), followed by afternoon (96.7%) and bedtime (95.2%), and lowest for morning prompts (94.3%). EMA adherence remained relatively stable across the 14-day monitoring period, with no marked decline over time.

Among 30 participants, 29 completed the follow-up survey. Most participants agreed or strongly agreed that the survey questions were easy to understand (89.7%) and that the survey system was easy to use (96.6%). Email and text reminders were rated as helpful by 96% and 100% of respondents, respectively. The median perceived burden of completing daily pain surveys was 1 (Interquartile Range [IQR]: 1–3) on a 0–10 scale, indicating low overall burden.

During the 14-day follow up, most variability in day-level average pain severity was attributable to between-person differences (between-person variance = 2.62; within-person variance = 0.64; Intraclass Correlation Coefficient [ICC] = 0.80). Similarly, the number of painful sites was largely between-person (between-person variance = 1.27; within-person variance = 0.17; ICC = 0.88), suggesting relatively stable multisite pain burden across the 14-day period.Most variability in total daily activity counts was attributable to between-person differences, with approximately 74% of variability reflecting stable differences across participants and 26% reflecting day-to-day fluctuations within individuals. In three-level variance decomposition models, with timepoints nested within days and days nested within individuals, pain severity exhibited substantial within-person variability. For pain severity, 64.7% of variance was attributable to between-person differences, 9.1% to day-to-day variability within individuals, and 26.2% to within-day variability across timepoints. For the number of painful sites, variance was primarily between-person (76.5%), with smaller day-to-day (5.3%) and within-day (18.2%) components. For interval-level PA counts, variability was dominated by within-day differences (84.5%), consistent with pronounced diurnal patterning; the day-level random effect was estimated at the boundary (≈0), indicating limited additional day-to-day clustering in this specification. Changes of pain severity and physical activity counts across day of week are presented in Supplementary Fig. 1. The changes in pain severity and physical activity counts by four time points across days are presented in Supplementary Fig. 2.

Association between momentary pain and physical activities for 14 days

Linear mixed effects models were used to examine the association between pain characteristics and PA counts across 4 time points per day. With only the linear term of PA across 4 time points considered, the association between overall pain severity and activity counts was statistically significant only at the wake-up time point (Fig. 3), where more severe pain was associated with fewer activity counts between wake-up and the morning pain survey period (β = −15,688, p = 0.045). We further added quadratic terms of time effects as the primary model (Table 4). Across models, the positive linear time coefficients and negative quadratic time coefficients indicated a nonlinear, inverted U-shaped diurnal PA pattern, with PA increasing from the early-day time points toward the middle of the day and decreasing later in the day. For overall pain severity, the changes in PA counts followed both a linear and a quadratic term (p < 0.001). The significant positive linear time term reflected an early-day increase in PA, while the significant negative quadratic term indicated that this increase slowed and reversed later in the day, which lead to a mid-day or afternoon peak followed by lower evening activity. After weekday/weekend status was added to the model, the significant three-way interactions among pain severity, weekday/weekend status, and both the linear and quadratic terms indicated that pain severity was associated with different diurnal PA trajectories on weekdays versus weekends (p < 0.04; Table 4A). Specifically, moderate-to-severe pain was associated with a lower-amplitude weekday PA pattern, whereas the weekend pattern showed a more pronounced mid-day peak among participants with moderate-to-severe pain (Fig. 4A). For pain presence at each site, shoulder, elbow, and back pain significantly modified the diurnal patterns of activities (Table 5). Shoulder and elbow pain were associated with a steeper early-day increase in PA and stronger negative curvature, reflecting a higher-amplitude pattern with a more pronounced mid-day peak and subsequent decline. In contrast, back pain was associated with a smaller early-day increase and weaker curvature, consistent with a flatter, lower-amplitude diurnal PA pattern (Fig. 4B). For pain severity at each pain site, pain severity in the elbow and hip was associated with diurnal changes in PA counts over a day (p for two-way interactions < 0.03; Table 6). Moderate-to-severe elbow pain was associated with a steeper early-day increase and greater negative curvature, suggesting a more pronounced mid-day PA peak, whereas moderate-to-severe hip pain was associated with a reduced early-day increase and less curvature, suggesting a flatter diurnal PA trajectory (Fig. 4C). We did not find significant three-way interactions with weekday/weekend status. For pain laterality, we found a significant two-way interaction between bilateral shoulder pain and quadratic time effects (p = 0.04; Table 7), suggesting that bilateral shoulder pain was associated with higher amplitude and higher peak activities during the day. With weekday/weekend status added to the model, we found significant three-way interactions for unilateral shoulder pain (p = 0.01). During weekends, older adults with unilateral shoulder pain tended to have a lower amplitude of PA patterns than those without shoulder pain (Fig. 4D).

Fig. 3.

Linear plots of predicted physical activity counts by overall pain severity across four time points. Lines represent model-estimated associations between momentary pain severity and predicted physical activity at the wake-up, morning, afternoon, and bedtime time points

Table 4.

Linear mixed effects models of ecological pain characteristics and physical activity count across 4 time points per day

Overall pain severity
	Pain severity (0–10)			Moderate-to-severe pain (≥ 4) vs no/mild pain
Fixed effects	β (× 1000)	SE	p-value	β (× 1000)	SE	p-value
Time effects only
Intercept	41.30	309.08	0.008	38.12	305.04	0.001
Pain severity	0.50	20.02	0.980	−0.51	79.26	0.995
Time	874.85	56.35	< 0.001	856.28	38.75	< 0.001
Time*Time	−190.66	10.71	< 0.001	−185.77	7.41	< 0.001
Pain severity*Time	−12.33	16.38	0.452	−49.67	69.66	0.476
Pain severity*Time*Time	3.43	3.13	0.274	14.94	13.66	0.274
AIC	44,713.05			44,703.94
BIC	44,777.64			44,768.53
Time and days effects
Intercept	−242.47	95.54	0.011	−234.60	75.39	0.002
Pain severity	10.96	22.73	0.630	71.86	90.92	0.429
Time	898.66	66.01	< 0.001	873.22	45.97	< 0.001
Time*Time	−196.41	12.63	< 0.001	−190.31	8.87	< 0.001
Weekend	49.83	135.17	0.712	6.72	88.94	0.940
Pain severity*Time	−24.03	18.90	0.204	−132.31	80.72	0.101
Pain severity*Time*Time	5.70	3.63	0.117	31.13	15.86	0.050
Pain severity*Weekend	−45.01	40.87	0.271	−308.55	173.38	0.075
Weekend*Time	−99.83	121.48	0.411	−56.29	81.33	0.489
Weekend*Time*Time	22.60	23.70	0.340	14.86	16.05	0.355
Pain severity*Weekend*Time	48.66	36.58	0.184	335.39	157.82	0.034
Pain severity*Weekend*Time*Time	−9.14	7.12	0.200	−63.96	30.97	0.039
AIC	44,597.53			44,575.39
BIC	44,694.36			44,672.21

Time of day is coded as 1, 2, 3, and 4, corresponding to the four ecological momentary assessment (EMA) prompt time points (wake-up, morning, afternoon, and bedtime) and the corresponding activity count intervals

AIC Akaike Information Criterion, BIC Bayesian Information Criterion

Bolded values indicate statistical signficant results (p<0.05)

Fig. 4.

Predicted physical activity counts across time of day by pain characteristics. Time of day is coded as 1 = wake-up, 2 = morning, 3 = afternoon, and 4 = bedtime ecological momentary assessment time points. A Overall pain severity by weekday/weekend status. B Pain presence at each site. C Pain severity at each site. D Shoulder pain laterality by weekday/weekend status

Table 5.

Linear mixed effects models of ecological pain characteristics and physical activity count across 4 time points per day

Pain vs. no pain at each pain site
	Shoulder pain			Elbow pain			Back pain
Fixed effects	β (× 1000)	SE	p-value	β (× 1000)	SE	p-value	β (× 1000)	SE	p-value
Time effects only
Intercept	96.93	301.44	0.748	69.23	302.74	0.819	−61.98	309.42	0.841
Pain (yes vs. no)	−131.91	91.48	0.150	−951.07	232.52	< 0.001	191.56	76.46	0.012
Time	798.45	39.13	< 0.001	811.92	33.05	< 0.001	902.47	42.44	< 0.001
Time*Time	−173.02	7.43	< 0.001	−175.54	6.22	< 0.001	−192.60	8.07	< 0.001
Pain (yes vs. no) *Time	140.08	70.85	0.048	1025.96	188.57	< 0.001	−144.75	65.07	0.026
Pain (yes vs. no) *Time*Time	−26.83	13.53	0.048	−199.14	36.89	< 0.001	26.91	12.70	0.034
AIC	44,702.46			44,667.49			44,700.77
BIC	44,767.06			44,732.09			44,765.36
Time and days effects
Intercept	105.30	310.22	0.734	56.19	311.94	0.857	−58.91	319.10	0.854
Pain (yes vs. no)	−179.13	94.90	0.059	−671.61	260.70	0.010	161.21	85.71	0.060
Time	770.67	45.61	< 0.001	806.61	38.35	< 0.001	890.78	50.45	< 0.001
Time*Time	−167.95	9.01	< 0.001	−175.17	7.57	< 0.001	−192.24	9.92	< 0.001
Weekend	−135.87	92.95	0.144	−47.03	78.67	0.550	−48.18	100.60	0.632
Pain (yes vs. no) *Time	192.64	83.43	0.021	811.71	230.17	< 0.001	−143.14	77.49	0.065
Pain (yes vs. no) *Time*Time	−38.65	16.41	0.019	−162.45	44.83	< 0.001	29.18	15.30	0.057
Pain (yes vs. no) *Weekend	227.88	171.24	0.184	−633.61	450.70	0.160	−37.55	160.02	0.815
Weekend*Time	90.42	85.55	0.291	12.30	71.80	0.864	6.02	91.51	0.948
Weekend*Time*Time	−14.10	16.90	0.404	1.10	14.14	0.938	2.44	18.02	0.892
Pain (yes vs. no) *Weekend*Time	−200.33	155.28	0.197	600.05	439.39	0.172	52.34	146.68	0.721
Pain (yes vs. no) *Weekend*Time*Time	39.71	30.54	0.194	−116.00	92.84	0.212	−9.48	28.93	0.743
AIC	44,579.09			44,538.07			44,579.83
BIC	44,675.91			44,634.89			44,676.65

Time of day is coded as 1, 2, 3, and 4, corresponding to the four ecological momentary assessment (EMA) prompt time points (wake-up, morning, afternoon, and bedtime) and the corresponding activity count intervals. Results for other pain sites are not presented because no statistically significant interactions between pain and time were observed

AIC Akaike Information Criterion, BIC Bayesian Information Criterion

Bolded values indicate statistical signficant results (p<0.05)

Table 6.

Linear mixed effects models of ecological pain characteristics and physical activity count across 4 time points per day

Pain severity at each pain site
	Moderate-to-severe pain (≥ 4) vs no/mild pain
	Elbow pain			Hip pain
Fixed effects	β (× 1000)	SE	p-value	β (× 1000)	SE	p-value
Time effects only
Intercept	−215.69	67.07	0.001	−263.12	68.82	0.001
Pain severity	−994.29	286.98	0.001	186.67	99.65	0.061
Time	825.59	32.98	< 0.001	871.37	35.50	< 0.001
Time*Time	−178.24	6.21	< 0.001	−187.61	6.72	< 0.001
Pain severity*Time	1002.31	257.68	< 0.001	−200.14	87.94	0.023
Pain severity*Time*Time	−196.15	52.66	< 0.001	42.69	17.43	0.014
AIC	44.684			44.70
BIC	44.75			44.76
Time and days effects
Intercept	−205.30	65.38	0.002	−240.58	68.26	< 0.001
Pain severity	−702.66	344.98	0.042	152.39	112.77	0.177
Time	818.26	38.27	< 0.001	859.52	41.60	< 0.001
Time*Time	−177.50	7.55	< 0.001	−186.82	8.19	< 0.001
Weekend	−53.01	78.47	0.499	−49.51	84.38	0.558
Pain severity*Time	784.78	321.49	0.015	−3.75	4.11	0.369
Pain severity*Time*Time	−162.31	64.53	0.012	146.66	51.20	0.008
Pain severity*Weekend	−62.97	724.63	0.931	−199.22	104.01	0.056
Weekend*Time	17.53	71.68	0.807	46.42	20.65	0.025
Weekend*Time*Time	0.04	14.12	0.998	−120.80	222.61	0.588
Pain severity*Weekend*Time	−238.49	803.90	0.767	13.52	77.04	0.861
Pain severity*Weekend*Time*Time	137.66	205.07	0.502	1.10	15.16	0.942
AIC	44,549.69			44,577.44
BIC	44,646.51			44,674.26

Time of day is coded as 1, 2, 3, and 4, corresponding to the four ecological momentary assessment (EMA) prompt time points (wake-up, morning, afternoon, and bedtime) and the corresponding activity count intervals. Results for other pain sites are not presented because no statistically significant interactions between pain and time were observed

AIC Akaike Information Criterion, BIC Bayesian Information Criterion

Bolded values indicate statistical signficant results (p<0.05)

Table 7.

Linear mixed effects models of ecological pain characteristics and physical activity count across 4 time points per day

Pain laterality at each pain site
	Shoulder pain
Fixed effects	β (× 1000)	SE	p-value
Time effects only
Intercept	26.21	313.52	0.933
Time	796.89	38.58	< 0.001
Time*Time	−172.03	7.62	< 0.001
Pain laterality
none	Ref
unilateral	−100.30	94.31	0.288
bilateral	−134.05	126.65	0.290
Pain laterality*Time
none	Ref
unilateral	97.09	81.65	0.235
bilateral	210.54	110.43	0.057
Pain laterality *Time*Time
none	Ref
unilateral	−18.70	16.01	0.243
bilateral	−44.76	21.83	0.041
Time and days effects
Intercept	66.09	314.95	0.834
Time	770.43	45.43	< 0.001
Time*Time	−167.88	8.98	< 0.001
Weekend	−136.33	92.60	0.141
Pain laterality
none	Ref
unilateral	−223.23	110.48	0.044
bilateral	−94.60	145.72	0.516
Pain laterality*Time
none	Ref
unilateral	228.21	96.86	0.019
bilateral	134.85	129.89	0.299
Pain laterality *Time*Time
none	Ref
unilateral	−45.01	18.99	0.018
bilateral	−29.25	25.70	0.255
Pain laterality *Weekend
none	Ref
unilateral	414.60	196.80	0.035
bilateral	−108.26	272.29	0.691
Weekend*Time	91.30	85.23	0.284
Weekend*Time*Time	−14.29	16.83	0.396
Pain laterality *Weekend*Time
none	Ref
unilateral	−438.04	178.70	0.014
bilateral	225.42	245.23	0.358
Pain laterality *Weekend*Time*Time
none	Ref
unilateral	87.78	35.07	0.012
bilateral	−45.79	48.42	0.345

Time of day is coded as 1, 2, 3, and 4, corresponding to the four ecological momentary assessment (EMA) prompt time points (wake-up, morning, afternoon, and bedtime) and the corresponding activity count intervals. Results for other pain sites are not presented because no statistically significant interactions between pain and time were observed

AIC Akaike Information Criterion, BIC Bayesian Information Criterion

Bolded values indicate statistical signficant results (p<0.05)

Discussion

In this pilot study of older adults with chronic multisite pain, we integrated EMA of pain characteristics with continuous accelerometer-derived PA metrics over 14 days. Overall, pain at specific sites, particularly elbow and back pain, was differentially associated with PA quantity, fragmentation, and diurnal activity patterns. Different pain characteristics were associated with diurnal PA patterns and contextual modulation of daily activity, particularly across time of day and weekday/weekend status. Taken together, these findings underscore the importance of characterizing daily activity in relation to pain within the context of time of day and weekday/weekend status. Given the observational pilot study design and small sample size, all findings should be interpreted cautiously as associations and not causal effects.

Baseline overall pain severity or pain interference were not significantly associated with average PA volume, activity fragmentation, or active minutes across the 14 days. Given the small sample size of the study, we are cautious to not overinterpret the findings. This may reinforce growing evidence that static pain measures may not fully characterize functional consequences of chronic pain in free-living environment. However, for pain at specific sites, back pain was consistently associated with lower total activity counts, fewer active minutes, and more fragmented activity patterns, as well as reduced activity during afternoon and evening hours. These findings align with prior evidence linking back pain to impaired postural control [35], lower reaction time [36], gait instability [37], fear-avoidance beliefs [38], all of which may disproportionately affect sustained or later-day activity. The diurnal pattern observed in this study suggests that individuals with back pain may have greater difficulty sustaining activity later in the day, as reflected by lower afternoon and evening activity. This pattern may be consistent with fatigue or pain exacerbation over the course of the day, which is supported by prior evidence linking chronic low back pain with fatigue and functional disability [39]. However, these baseline associations provided only partial insight into the complexity of pain-activity relationship. Future studies should further explore the pathways linking back pain and diurnal patterns of activity.

Across all EMA models, PA exhibited a robust nonlinear diurnal pattern, characterized by an early-day increase, a mid-day peak, and a decline toward evening. We found that higher momentary pain severity was associated with lower PA after wake-up, with attenuation later in the day, and relatively higher activity at night, suggesting that pain may exert its strongest influence on activity early in the daily activity cycle and may also relate to disrupted sleep. Our findings are consistent with prior work demonstrating time-of-day variation in pain-activity coupling. For instance, Zhaoyang and colleagues showed that on mornings when individuals catastrophized more about their pain, they subsequently engaged in less MVPA and more sedentary behavior throughout the day [40]. Another study using EMA of pain found that higher momentary pain was associated with lower PA specifically during morning hours, particularly among individuals with high pain-related activity interference [41]. One potential explanation is that pain co-occurs with other symptoms such as morning joint stiffness, a hallmark feature of osteoarthritis and other musculoskeletal conditions that may transiently limit mobility upon waking. The higher activity associated with more severe pain in our study may indicate poorer sleep with older adults with pain. Previous studies without momentary pain assessment demonstrated that older adults with higher pain ratings had poorer sleep quality [42, 43], and this impact is mediated by anxiety symptoms [43]. Taken together, these findings highlight that pain-activity relationships are strongly time-dependent and likely shaped by both physiological symptoms and behavioral adaptations across the day.

A key novel finding of this study is that pain severity modified diurnal PA patterns differently on weekdays versus weekends, as demonstrated by significant three-way interactions among pain severity, time of day, and weekday/weekend status. Specifically, individuals with moderate-to-severe pain exhibited higher mid-day PA on weekdays but lower mid-day PA on weekends compared with those with no or mild pain. Prior EMA studies have also shown that pain–activity relationships are highly context-dependent and vary within individuals across time, rather than reflecting stable between-person differences alone [40, 41]. On weekdays, externally imposed routines and obligations (e.g., caregiving tasks, household responsibilities, scheduled activities) may compel individuals with higher pain to remain active during mid-day hours despite greater symptom burden, consistent with models of activity persistence and overactivity in chronic pain [44, 45]. In contrast, weekends may allow greater flexibility in pacing and activity timing, enabling individuals with higher pain to down-regulate mid-day activity, while those with lower pain maintain or increase activity during this period. This interpretation is consistent with prior work demonstrating that daily routines and role demands strongly structure PA patterns and may constrain the ability to self-regulate activity in response to symptoms [46, 47]. Collectively, these findings highlight the importance of EMA-informed and time-specific modeling approaches for capturing dynamic and context-dependent pain–activity relationships in daily life.

Pain severity at specific locations was associated with diurnal PA trajectories, particularly for elbow and hip pain. Higher elbow pain severity was associated with a steeper early-day increase and a higher mid-day activity peak, followed by a sharper decline later in the day. One possible explanation is that elbow pain in older adults may reflect overuse or repetitive upper-extremity activity rather than generalized mobility limitation. Individuals who engage more frequently in upper-extremity intensive tasks (e.g., household chores or caregiving) may accumulate mechanical load, resulting in pain while maintaining relatively high overall activity levels. Prior studies also suggest that upper extremity pain may not be detrimental to activity performance in the same way that lower extremity pain is [22]. Importantly, elbow pain may have less impact on ambulation and lower-extremity function compared to back or hip pain, allowing older adults to maintain typical daily movement patterns [22]. Finally, elbow pain may serve as a marker of a highly active subgroup rather than a causal driver of increased activity. In this context, the observed association likely reflects underlying behavioral patterns, where more active individuals are both more exposed to repetitive strain and more likely to report upper-extremity pain.

Pain laterality was also associated with activity patterns, particularly for elbow and ankle/foot/toe pain. Participants with unilateral or bilateral elbow pain demonstrated greater afternoon activity compared to those without elbow pain, while unilateral ankle/foot/toe pain was associated with increased morning and evening activity. These findings may reflect differences in participant characteristics, activity routines, or activity exposure rather than a direct effect of pain laterality on PA. Unilateral pain may allow individuals to compensate using the contralateral limb, maintaining activity at certain times, whereas bilateral pain may impose more global constraints. This interpretation is consistent with prior evidence indicating that bilateral pain conditions are associated with greater difficulty in performing PA, as common tasks such as walking and stair climbing that require coordinated use of both limbs [48]. Our previous work similarly observed greater afternoon activity among older adults with unilateral hand/wrist pain, but lower activity levels in those with unilateral knee pain [22]. Evidence regarding greater activity in older adults with unilateral foot pain remains limited, and this finding was based on baseline pain characteristics and averaged PA across 14 days. In contrast, we did not observe significant associations between lower extremity pain laterality and PA in the EMA-based analyses. Therefore, further studies are needed to clarify the temporality and directionality of the relationship between pain laterality and PA, particularly for lower extremity pain.

Although the EMA models primarily focused on PA volume (e.g., activity counts), the observed associations with activity fragmentation are noteworthy. Higher ASTP among participants with back pain indicates more fragmented activity patterns, which have been linked to fatigability, lower physical function, and mortality in older adults [33, 49]. Fragmentation of daily activities may reflect difficulty sustaining prolonged activity due to pain-related discomfort, neuromuscular inefficiency, or cardiorespiratory weakness [33]. In contrast, lower ASTP among participants with elbow pain suggests longer, more continuous activity bouts, consistent with their higher overall activity levels. These findings support the conceptualization of PA as a complex and regulated system that may become dysregulated in the presence of certain pain conditions. Future work leveraging PA complexity metrics, such as multiscale entropy [50], in conjunction with time-varying pain measures may further elucidate how chronic pain alters the adaptive capacity of daily movement patterns.

Several limitations should be acknowledged. First, the small sample size limits statistical power and precludes extensive multivariable adjustment. Although we used parsimonious models and conducted age- and sex-adjusted sensitivity analyses, residual confounding may impact the results. Potential confounders include comorbidities including osteoarthritis, depression, fear of falling, physical function, analgesic use, caregiving or occupational roles, and other behavioral or environmental factors that may influence both pain reporting and PA. Therefore, the counterintuitive positive associations observed for elbow and ankle/foot/toe pain should be interpreted cautiously and may reflect characteristics of participants who reported pain at those sites rather than site-specific causal effects on PA. Second, the sample was predominantly White and highly educated, which may limit generalizability to more diverse populations. Third, although wrist-worn accelerometers capture overall movement volume well, they may be less sensitive to certain postural transitions or lower-extremity loading, which are particularly relevant for musculoskeletal pain. Future longitudinal studies with larger sample sizes are warranted to integrate EMA approaches with wearable sensors to elucidate the temporal and potentially causal relationships between pain fluctuations and daily movement behaviors.

In conclusion, this study highlights heterogeneity in the relationships between chronic pain and daily PA among older adults. Overall pain severity alone may not be sufficient to explain differences in PA volume and patterns. Instead, pain location, laterality, and time of day may play critical roles in understanding the complex relationships between musculoskeletal pain and daily activity behaviors. These findings support a shift toward more granular assessment of pain and activity in aging research. Future larger-scale studies should integrate EMA-derived pain phenotypes with advanced PA metrics, including complexity, to identify temporal and causal pathways that may inform personalized interventions aimed at promoting daily activities and maintaining physical functioning in older adults with chronic pain.

Abbreviations

PA

Physical activity

PRIME

Pain and Activity Rhythms In Multisite-pain Elders

MVPA

Moderate-to-vigorous physical activity

TAC

Total activity counts

ASTP

Active-to-sedentary transition probability

BMI

Body mass index

BPI

Brief Pain Inventory

CESD-R

Center for Epidemiologic Studies Depression Scale–Revised

SPPB

Short Physical Performance Battery

CPIW

CentrePoint Insight Watch

EMA

Ecological momentary assessment

FES-I

Falls Efficacy Scale–International

ICC

Intraclass correlation coefficient

IQR

Interquartile range

MoCA

Montreal Cognitive Assessment

REML

Restricted maximum likelihood

SD

Standard deviation

Authors’ contributions

YC designed the study, secured funding, collected the data, conducted statistical analyses, interpreted the data, and drafted the manuscript. MM conducted statistical analyses, prepared tables and figures, and interpreted the findings. PS conducted statistical analyses and interpreted the data. JH, SH and AB collected the data. All authors critically reviewed and approved the final manuscript.

Funding

This work was supported by the University of Pittsburgh School of Nursing Research Catalyst Award and the Aging and Gerontological Nursing Research (AGE) HUB.

Data availability

Data are available from the corresponding author upon reasonable request, subject to applicable institutional approvals and participant privacy protections.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board of the University of Pittsburgh. All participants provided written informed consent prior to participation. All procedures involving human participants/human data were performed in accordance with relevant guidelines and regulations, including the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interest

The authors declare no competing interests.