Continuous preoperative activity patterns measured by wearable sensors are associated with recovery after total knee arthroplasty

Abstract

Objective

Physical activity is increasingly recognized as an important factor in recovery after orthopedic surgery, but objective evidence remains limited. This study examined whether wearable-derived preoperative activity levels were associated with recovery outcomes following Total Knee Arthroplasty (TKA).

Methods

In this prospective longitudinal study, 30 adults undergoing unilateral TKA for knee osteoarthritis wore validated thigh-mounted accelerometers for up to two weeks preoperatively and for three months postoperatively. Participants were classified into low-, moderate-, and high-activity groups based on preoperative 24-hour activity intensity profiles. Recovery was assessed using continuous accelerometer-derived activity metrics and patient-reported outcomes (PROMs), including the Knee injury and Osteoarthritis Outcome Score (KOOS).

Results

Higher preoperative activity levels were associated with more favorable objective recovery after TKA. Compared with the low-activity group, the high-activity group showed higher postoperative activity intensity and total steps, with 2.6-fold higher activity intensity and 2.2-fold higher total steps. During the first postoperative month, the high-activity group also showed faster increases in activity intensity and total steps. In contrast, KOOS subscale scores improved over time but did not differ significantly between activity groups.

Conclusion

The findings of this study highlight the importance of accounting for baseline activity when interpreting wearable-derived postoperative recovery patterns. They also suggest that objective monitoring can provide information on functional recovery that may not be captured by PROMs.

Keywords

wearable sensors total knee arthroplasty physical activity patterns recovery trajectories rehabilitation

1. Introduction

Total knee arthroplasty (TKA) is a commonly performed procedure designed to relieve pain and restore function in patients with advanced knee osteoarthritis unresponsive to conservative treatment.¹ Although TKA is highly successful overall, recovery trajectories vary considerably among individuals, reflecting the influence of multiple factors such as perioperative care, pain management, and rehabilitation strategies.^2,3 Understanding these factors is essential for optimizing patient management, tailoring rehabilitation, and ultimately improving satisfaction and functional recovery.⁴

Preoperative physical activity has been identified as a potentially adjustable factor associated with postoperative recovery.^5–9 However, most prior studies investigating this relationship have relied on discrete assessments; either subjective, such as questionnaires, or objective, like accelerometer readings taken during brief laboratory visits.^5,6,10 Such discrete assessments may limit the ability to capture habitual and nuanced activity patterns.¹¹ Moreover, earlier research has predominantly relied on patient-reported outcome measures (PROMs) such as the Oxford Knee Score or Knee Society Score, which may overestimate recovery when compared with objective performance-based assessments.¹²

Advances in wearable sensor technology now enable continuous monitoring of daily physical activity, providing detailed insights into real-world behaviors that go beyond self-reports.¹³ Diurnal activity profiles captured by wearable sensors can reflect patients’ functional capacity and behavioral adaptations to chronic pain, making them potential markers of postoperative recovery patterns.^14,15 However, most existing research has focused on postoperative activity, leaving a critical gap in understanding the relationship between preoperative activity patterns and postoperative functional and PROMS.¹⁶

This study investigated whether preoperative physical activity patterns, measured objectively using wearable sensors, are associated with recovery after TKA. Specifically, we examined how daily activity profiles before surgery relate to postoperative mobility, pain, and quality of life. We hypothesized that patients with more intense and consistent preoperative activity would show more favourable objective recovery after surgery.

2. Method

2.1. Study design and participants

This prospective longitudinal observational study was conducted at Aalborg University Hospital and Capio Private Hospital in Denmark between March and October 2024. Ethical approval was granted by the North Denmark Regional Committee on Health Research Ethics (Journal number: N-20230035). All study procedures complied with the Declaration of Helsinki and the General Data Protection Regulation. Written informed consent was obtained from all participants prior to enrollment.

Participants were eligible for inclusion if they were adults aged 18 years or older and scheduled for unilateral TKA due to knee osteoarthritis that had not responded to non-surgical management. Inclusion criteria required independent ambulation, ownership of a smartphone compatible with the study’s mobile application, and willingness to wear a physical activity tracker continuously throughout the study period. Participants were excluded if they had a Clinical Frailty Scale score of 5 or higher, severe comorbidities, recent surgery involving the lower limbs or spine within the past six months, cognitive impairment, dermatological conditions at the sensor attachment site, or used walking aids prior to surgery. Perioperative management was standardized across both sites, including analgesia, postoperative mobilization, and rehabilitation. Patients at both hospitals followed the same clinical pathway, with the same postoperative treatment and rehabilitation principles.

2.2. Physical activity data

Participants were equipped with SENS Motion® trackers (SENS Innovation ApS), which are medical-grade wearable accelerometers that have been validated in previous studies.^17,18 The sensor was attached to the lateral distal thigh and worn continuously, 24 hours per day, throughout the monitoring periods. Data were synchronized daily via a dedicated smartphone application to a secure cloud platform. The sensor battery supported uninterrupted monitoring across the 3-month follow-up period, and participants were instructed to open the mobile application regularly to enable synchronization. If synchronization had not occurred within 72 hours, participants were contacted directly. Patch or sensor replacement was performed as needed during follow-up.

Data collection began two weeks prior to surgery and continued for three months postoperatively. Preoperative physical activity was assessed using available valid sensor data collected within an up-to-14-day monitoring window before surgery. This window was selected to provide stable and reliable estimates of habitual physical activity, as recent evidence indicates that 7 to 10 days of wearable monitoring are generally required to achieve acceptable reliability for key activity metrics.¹⁹ A day was considered valid/analyzable when at least 95% of the 24-hour recording contained valid sensor data. Periods without valid sensor signal or synchronization were treated as missing/non-wear and were excluded from daily summaries.

The SENS Motion algorithm classified daily physical activity into several physical activity parameters.

• Activity intensity count: A measure of daily movement intensity, calculated as the average magnitude of acceleration in each 5-second epoch across the entire recorded signal for each day.

• Step count: Total number of daily steps categorized by walking type (continuous regular walking, sporadic walking, and slow continuous walking).

• Sit-to-stand transitions: Daily count of transitions from a seated to a standing position.

• Postural allocation: Percentage of daily time spent in different body positions, including lying, sitting, standing, and walking.

Physical activity parameters were aggregated into 24-hour summaries per participants and segmented into different periods (preoperative, and postoperative months 1, 2, and 3). In addition, hourly activity intensity profiles were generated by summing intensity counts within each hour of the day and averaging across days per participant and period. This enabled the construction of individualized diurnal intensity curves for each recovery stage.

Preoperative hourly physical activity intensity counts were analyzed to identify distinct patterns of daily activity behavior. For each participant, a mean 24-hour preoperative activity profile was created by averaging their daily activity intensity curves over all monitored days before surgery. This mean profile was then modeled as a smooth curve using a 7-term Fourier basis expansion, representing the typical diurnal activity rhythm for that participant as a sum of sinusoidal functions. This approach acts as a low-pass filter and effectively captures the dominant cyclical patterns in daily activity while minimizing high-frequency noise and smoothing out minor day-to-day fluctuations. The resulting functional representations were normalized using feature-wise z-scores and clustered using hierarchical clustering.²⁰ Candidate solutions with k = 2, 3, and 4 were compared using internal validity indices (silhouette, Calinski–Harabasz, and Davies–Bouldin). The 3-cluster solution was retained as the primary specification based on overall interpretability and sample-size considerations.

Recovery across the resulting preoperative activity clusters was then examined by visually comparing hourly activity intensity profiles at 1, 2, and 3 months after surgery. In addition, daily values for step counts, including regular continuous, slow continuous, sporadic walking, and total steps calculated as the sum of these three categories, were extracted together with sit-to-stand transitions, activity intensity count, and time spent in distinct postural behaviors including resting, sitting, standing, and walking. These outcomes were analyzed across the 90-day postoperative period, as described in the Statistical Analysis section.

2.3. Patient-reported outcomes (PROMs)

Patient-reported outcomes (PROMs) were collected electronically using REDCap (Research Electronic Data Capture).²¹ The Knee Injury and Osteoarthritis Outcome Score (KOOS) was administered at each study phase, including the preoperative assessment and the postoperative assessments at 1, 2, and 3 months. The KOOS comprises five subscales: Pain, Symptoms, Activities of Daily Living (ADL), Sport and Recreation, and Quality of Life (QoL).

2.4. Statistical analysis

Baseline characteristics were compared across the three preoperative activity clusters using one-way analysis of variance (ANOVA) or Fisher’s exact test, as appropriate.

The primary outcome was activity intensity count, derived from the wearable sensor as a measure of overall movement intensity based on acceleration magnitude. Secondary outcomes included step-based measures, sit-to-stand transitions, postural behavior outcomes, and KOOS subscales.

Postoperative physical activity outcomes were analyzed using generalized estimating equations (GEE) to account for repeated daily observations within participants. Analyses were performed within three predefined postoperative periods (days 0–30, 31–60, and 61–90). Within each period, cluster-only models were used to estimate between-cluster differences, and additional models including day-in-block, cluster, and their interaction were used to examine differences in recovery slopes. Negative binomial GEE models were used for step-based outcomes and sit-to-stand transitions, Gamma GEE models with a log link were used for activity intensity, and Gaussian GEE models on logit-transformed proportions were used for postural behavior outcomes. For negative binomial and Gamma GEE models, exponentiated coefficients are presented, whereas Gaussian GEE estimates are reported on the transformed model scale. False discovery rate adjustment was applied to key between-cluster comparisons to address multiple testing. Model assumptions were assessed using residual and distributional diagnostics. Within each postoperative month, cluster-specific rates of change were estimated from the fitted models, and pairwise comparisons were used to assess whether these rates differed between clusters. Results are reported as model-based effect estimates with 95% confidence intervals.

For visualization, daily wearable-derived recovery outcomes were plotted as model-based estimated trajectories over the first 90 postoperative days with 95% confidence intervals.

KOOS subscales were analyzed using linear mixed-effects models with fixed effects for time, preoperative activity cluster, and their interaction, with a random intercept for participant. For visualization, KOOS subscale scores were plotted as group means with 95% confidence intervals.

To assess potential baseline confounding, key baseline variables were examined across clusters, and additional sensitivity analyses were performed with adjustment for BMI, age, sex, and comorbidity presence.

Statistical significance was defined as a two-sided p value < 0.05. Missing data were not imputed, and longitudinal analyses were performed using all available observations. All analyses were conducted in Python version 3.12.2.

3. Results

3.1. Study population

Baseline demographic and clinical characteristics stratified by preoperative activity cluster are presented in Table 1. The study cohort comprised 30 participants, including 17 females and 13 males, distributed across the low-activity (n = 10), moderate-activity (n = 8), and high-activity (n = 12) groups. The mean age was 68.1 years (SD = 9.9), with participants ranging from 44.7 to 86.6 years of age. The mean body mass index (BMI) was 29.6 kg/m² (SD = 5.0), ranging from 21.8 to 39.1 kg/m².

Table 1.

Baseline demographic and clinical characteristics stratified by preoperative activity.

Variables				Low (n=10)	Moderate. (n=8)	High (n=12)
Sex n (%)	Male			5 (50%)	4 (50%)	4 (33%)
Sex n (%)	Female			5 (50%)	4 (50%)	8 (67%)
Age Mean ± SD				68.8 ± 7.4	69.4 ± 11.9	66.6 ± 10.9
Height Mean ± SD				172 ± 9.5	172 ± 9.4	173 ± 11.2
Weight Mean ± SD				93.9 ± 18.9	85.9 ± 19.6	84.8 ± 14.2
BMI Mean ± SD				31.7 ± 6.1	28.8 ± 4.7	28.3 ± 3.9
Surgery side n (%)	Right			6 (60%)	4 (50%)	6 (50%)
Surgery side n (%)	Left			4 (40%)	4 (50%)	6 (50%)
Surgery history n (%)	No			4 (40%)	5 (63%)	3 (25%)
	Yes	Type of previous surgery	Arthroplasty	2 (20%)	0	2 (17%)
			Osteotomy	1 (10%)	0	0
			Soft tissue	3 (30%)	3 (38%)	7 (58%)
		Side of previous surgery	Ipsilateral	4 (40%)	1 (13%)	4 (33%)
		Side of previous surgery	Contralateral	2 (20%)	2 (25%)	4 (33%)
		Location of previous surgery	Spine	0	0	1 (8%)
			Hip	0	0	1 (8%)
			Thigh	0	0	1 (8%)
			Knee	6 (60%)	3 (38%)	6 (50%)
Comorbidities	No			3 (30%)	4 (50%)	6 (50%)
	Hypertension			1 (10%)	1 (13%)	4 (33%)
	Heart disease			2 (20%)	0	1 (8%)
	Chronic pain syndromes			2 (20%)	1 (13%)	0
	Diabetes			2 (20%)	0	0
	Depression			0	1 (13%)	0
	COPD			0	1 (13%)	0
	Previous cancer			0	0	1 (8%)

Sixteen participants underwent right-sided, and 14 underwent left-sided TKA. Previous surgery was reported by 18 participants, most commonly involving the knee, and 17 participants reported at least one comorbid condition, most commonly hypertension. Baseline characteristics were broadly similar across clusters; BMI was the only variable that differed significantly between groups (p = 0.040), with lower BMI in the moderate-activity group than in the low-activity group (post hoc p = 0.043). In sensitivity analyses adjusting for BMI, age, sex, and comorbidity presence, the principal associations for activity intensity, total steps, and regular continuous steps remained broadly consistent. Some secondary step-pattern associations, including sporadic steps and the moderate-versus-low comparison for slow continuous steps, were attenuated after adjustment and were therefore interpreted cautiously.

3.2. Physical activity outcomes

Wearable data completeness was high throughout data collection. Across all recorded study days, mean valid wear percentage was 95.3%, and 2659 of 2794 recorded days (95.2%) met the analyzable-day criterion. In the preoperative period, 299 of 378 expected days were available, of which 297 days (99.3% of recorded preoperative days) were analyzable. By study period, analyzable days represented 98.6% of expected days in month 1, 100.0% in month 2, and 89.8% in month 3, since the three participants had incomplete postoperative activity records, with data collection ending at postoperative days 48, 53, and 72, respectively. These individuals were included in the analyses up to their last available time point.

Analysis of preoperative physical activity data identified three distinct patterns using hierarchical clustering of 24-hour activity intensity profiles, revealing clear inter-individual differences in diurnal activity patterns prior to TKA (Figure 1). Cluster 1, characterized by low activity, showed the lowest levels of physical activity throughout the day, with only a gradual increase beginning in the early morning and remaining consistently low across waking hours. Cluster 2, the moderate activity group, displayed a pronounced peak in the morning followed by a steady decline during the afternoon, with a smaller secondary rise later in the day. Cluster 3, representing high activity, maintained the highest and most sustained activity levels across the day, with sharp increases in the early morning that remained elevated into the late afternoon before gradually tapering in the evening.

Figure 1.

Preoperative hourly activity intensity profiles by preoperative activity group. Lines represent mean hourly activity intensity counts across the 24-hour day for each preoperative activity group. Shaded bands indicate 95% confidence intervals. Orange indicates the low-activity group, blue the moderate-activity group, and green the high-activity group.

Over the three-month postoperative period, distinct recovery trajectories were observed across the preoperative activity clusters (Figure 2). All groups experienced an initial decline in physical activity immediately after surgery. The high-activity group appeared to show the greatest recovery in both activity volume and diurnal rhythmicity, while the moderate-activity group showed intermediate recovery with progressive increases in activity and gradual re-establishment of a structured daily pattern. In contrast, the low-activity group exhibited only modest gains, maintaining a blunted and low-intensity activity profile throughout the follow-up period.

Figure 2.

Postoperative hourly activity intensity profiles by preoperative activity group. Panels show mean hourly activity intensity counts for each preoperative activity group during postoperative months 1, 2, and 3. Shaded bands indicate 95% confidence intervals. Orange indicates the low-activity group, blue the moderate-activity group, and green the high-activity group.

Significant between-cluster differences were observed for several postoperative activity outcomes (Table 2, Figure 3). Compared with the low-activity cluster, the moderate- and high-activity clusters showed higher activity intensity, total steps, regular continuous steps, and slow continuous steps, with the largest effects observed for activity intensity and total steps. Sporadic steps were also higher in the high-activity cluster than in the low-activity cluster. Sit-to-stand transitions did not differ significantly between clusters after correction for multiple testing.

Table 2.

Key longitudinal associations between preoperative activity cluster and postoperative physical activity outcomes. Effect estimates are exponentiated coefficients for negative binomial and Gamma GEE models and regression coefficients on the transformed model scale for Gaussian GEE models.

Outcome	Model	Comparison	Effect estimate (95% CI)	p-value	FDR-adjusted p
Activity intensity	Gamma GEE	Moderate vs Low	1.97 (1.45–2.67)	<0.001	<0.001
		High vs Low	2.60 (2.04–3.31)	<0.001	<0.001
		High vs Moderate	1.32 (1.02–1.71)	0.037	0.079
Total steps	NB GEE	Moderate vs Low	1.72 (1.26–2.37)	<0.001	0.003
		High vs Low	2.19 (1.70–2.83)	<0.001	<0.001
		High vs Moderate	1.27 (0.96–1.69)	0.096	0.175
Steps	NB GEE	Moderate vs Low	2.30 (1.41–3.76)	<0.001	0.003
		High vs Low	2.96 (1.98–4.41)	<0.001	<0.001
		High vs Moderate	1.29 (0.86–1.93)	0.225	0.274
Slow continuous steps	NB GEE	Moderate vs Low	1.63 (1.13–2.37)	0.010	0.024
		High vs Low	1.98 (1.46–2.68)	<0.001	<0.001
		High vs Moderate	1.21 (0.89–1.65)	0.228	0.274
Sporadic steps	NB GEE	Moderate vs Low	1.45 (0.99–2.13)	0.057	0.115
		High vs Low	1.79 (1.23–2.60)	0.002	0.007
		High vs Moderate	1.23 (0.92–1.65)	0.163	0.244
Sit-to-stand transitions	NB GEE	Moderate vs Low	0.90 (0.62–1.30)	0.564	0.604
		High vs Low	1.05 (0.74–1.49)	0.800	0.828
		High vs Moderate	1.17 (0.97–1.40)	0.099	0.175
Resting time	Gaussian GEE	Moderate vs Low	-0.27 (-0.70–0.16)	0.226	0.274
		High vs Low	-0.47 (-0.87–-0.08)	0.019	0.044
		High vs Moderate	-0.21 (-0.48–0.07)	0.138	0.217
Sitting time	Gaussian GEE	Moderate vs Low	0.78 (0.37–1.18)	<0.001	<0.001
		High vs Low	1.00 (0.60–1.40)	<0.001	<0.001
		High vs Moderate	0.23 (-0.14–0.59)	0.224	0.274
Standing time	Gaussian GEE	Moderate vs Low	-0.32 (-1.01–0.38)	0.370	0.411
		High vs Low	-0.34 (-0.95–0.27)	0.278	0.321
		High vs Moderate	-0.02 (-0.52–0.48)	0.930	0.930
Walking time	Gaussian GEE	Moderate vs Low	0.30 (-0.07–0.67)	0.117	0.195
		High vs Low	0.57 (0.14–1.00)	0.009	0.024
		High vs Moderate	0.27 (-0.16–0.71)	0.214	0.274

Figure 3.

Postoperative recovery of daily activity outcomes by preoperative activity group. Panels show model-based estimated trajectories across the first 90 postoperative days; shaded bands indicate 95% confidence intervals. Points represent observed daily values, dashed horizontal lines show group-specific preoperative means, and the dashed vertical line marks surgery. Y-axes were limited for visualization; model estimates used all valid observations. Orange indicates the low-activity group, blue the moderate-activity group, and green the high-activity group.

For postural behavior outcomes, fewer cluster-related differences were observed (Table 2, Figure 4). Resting time was lower in the high-activity cluster than in the low-activity cluster, sitting time was higher in both the moderate- and high-activity clusters than in the low-activity cluster, and walking time was higher in the high-activity cluster than in the low-activity cluster. Standing time did not differ significantly between clusters after correction for multiple testing.

Figure 4.

Postoperative recovery of daily postural behavior outcomes by preoperative activity group. Panels show model-based estimated trajectories across the first 90 postoperative days; shaded bands indicate 95% confidence intervals. Points represent observed daily values, dashed horizontal lines show group-specific preoperative means, and the dashed vertical line marks surgery. Y-axes were limited for visualization; model estimates used all valid observations. Orange indicates the low-activity group, blue the moderate-activity group, and green the high-activity group.

Time-by-cluster analyses identified significant differences in rates of change for selected outcomes. During the first postoperative month, activity intensity increased faster in the high-activity cluster than in the low-activity cluster (effect ratio per day 1.02, 95% CI 1.00 to 1.03; p = 0.016). Total steps also increased faster in the high-activity cluster than in the low-activity cluster during month 1, although this difference was smaller (effect ratio per day 1.01, 95% CI 1.00 to 1.03; p = 0.049). Additional differences were observed for sporadic steps in the third postoperative month, but these later findings were interpreted cautiously because of greater instability in late follow-up observations.

3.3. PROMs: KOOS subscales

Figure 5 shows KOOS subscale scores over time with 95% confidence intervals. No significant between-cluster differences were observed for any KOOS subscale at baseline or across follow-up (all p > 0.05). Postoperatively, all groups demonstrated improvements across KOOS domains, with the most pronounced gains observed in Pain, Sport and Recreation, and Quality of Life (QoL) scores by week 12 (p < 0.05). While the Symptoms and ADL subscales also improved over time, these changes were not statistically significant.

Figure 5.

KOOS subscale scores over time by preoperative activity group. Panels show mean KOOS scores for Pain, Symptoms, Activities of Daily Living (ADL), Sport and Recreation, and Quality of Life (QoL) from the preoperative assessment to 3 months postoperatively. Error bars indicate 95% confidence intervals. Orange indicates the low-activity group, blue the moderate-activity group, and green the high-activity group.

4. Discussion

In this study, preoperative activity clusters identified from continuous wearable sensor data were associated with distinct activity patterns of postoperative recovery following TKA. Patients in the high-activity cluster demonstrated more favorable objective recovery trajectories, including higher postoperative activity levels overall and a faster early rate of recovery, particularly in activity intensity and total steps during the first postoperative month. These findings support the relevance of continuously measured preoperative activity behavior as an indicator of early postoperative recovery and highlight the potential value of wearable monitoring for characterizing recovery patterns after TKA.

It has been observed that participants with higher levels of preoperative physical activity experienced better recovery after TKA, characterized by faster functional improvement and an earlier return to baseline activity. This observation aligns with Kennedy et al.'s findings, which suggest that greater preoperative function predicts better postoperative outcomes on performance tests.⁷Similarly, Oka et al. found that increased preoperative sedentary behavior was associated with poorer Knee Society Scores.⁵ However, not all studies establish a direct link between preoperative physical activity and postoperative recovery. For instance, Boersma et al. used the self-reported SQUASH questionnaire to assess preoperative activity and found no association with time to return to work, suggesting that occupational and psychosocial factors may play a greater role in postoperative recovery.²² Rooks et al. conducted a randomized controlled trial of prehabilitation in patients awaiting TKA and found that while preoperative exercise improved surgical readiness, it did not translate into long-term functional gains.²³ Their study population consisted mainly of relatively active individuals, indicating that prehabilitation effects might differ in patients with lower baseline activity levels. Onerup et al. found that physically active colorectal surgery patients reported better recovery, although objective measures showed no difference.²⁴ Vasta et al.'s systematic review of TKA and THA prehabilitation revealed mixed outcomes with no consistent functional improvements.¹⁰ These findings underscore the importance of assessing habitual physical activity to better understand recovery outcomes.

Postoperative gains in structured ambulatory behavior, specifically regular and slow continuous walking, were most pronounced in participants with higher preoperative activity levels. In contrast, sporadic walking remained unchanged, indicating recovery is driven by improvements in sustained ambulation. These findings align with Christensen et al.²⁵ and Bin Sheeha et al.,²⁶ who found step count and moderate-to-vigorous activity improved postoperatively, while low-effort behaviors were less responsive. Sit-to-stand transitions increased across groups without significant cluster variation, suggesting this function relates more to general recovery. Activity intensity increased particularly among moderate and high activity clusters, supporting evidence that higher baseline function leads to faster recovery of movement.⁷ We observed decreased resting time and increased sitting and standing durations, with greater walking gains in higher-activity participants, supporting a recovery model progressing from rest to upright postures before dynamic activity. These patterns align with Hammett et al.,¹⁶ Harding et al.,²⁷ and Oka et al.,⁵ showing sedentary behavior persists despite improved pain and function.

Despite clear differences in objectively measured recovery, PROMs (KOOS subscales) did not significantly differ across preoperative physical activity clusters. In the Symptoms subscale, the low-activity cluster appeared numerically higher across follow-up; however, this separation was already present preoperatively and did not widen over time. Together with the wide uncertainty intervals and the non-significant cluster and time-by-cluster terms, this suggests baseline variation rather than a true cluster-related difference in postoperative symptom recovery. All participants reported improvements in pain, sport and recreation, and quality of life by 12 weeks, while gains in symptoms and activities of daily living were more modest. These findings echo prior research showing a frequent mismatch between PROMs and objective metrics. For example, Kennedy et al.⁷ found that self-reported tools like the WOMAC failed to detect early postoperative declines captured by objective tests. Similarly, Hammett et al.¹⁶ and Bin Sheeha et al.²⁶ Observed that improvements in self-reported pain and function were not accompanied by corresponding increases in physical activity, suggesting that PROMs may reflect perceived relief more than actual behavioral change. Ceiling effects in tools like KOOS may further obscure subtle differences in recovery trajectories. The consistency of PROM improvement across clusters in our study, despite divergent objective gains, reinforces the value of combining subjective and objective assessment as mentioned in literature.³ Such multidimensional frameworks are essential for capturing the full complexity of recovery and optimizing individualized care after TKA.

The findings of this study have important clinical and research implications. Identifying distinct preoperative physical activity patterns through wearable sensor data may offer valuable opportunities to inform perioperative care for TKA patients. Stratifying patients by activity patterns may help inform future rehabilitation planning and recovery monitoring. Integrating wearable-based activity monitoring into clinical practice could also provide continuous, objective feedback on patient progress, enabling timely interventions when deviations from expected recovery occur. Such approaches may ultimately support recovery monitoring and help reduce the proportion of patients, estimated at around 20%, who remain dissatisfied or experience limited benefit after TKA.¹ Moreover, these findings highlight opportunities for targeted intervention. Patients with low baseline activity may particularly benefit from prehabilitation programs or individualized perioperative education and rehabilitation strategies. Evidence from systematic reviews suggests that prehabilitation can enhance preoperative functional capacity and may shorten early recovery time in selected orthopedic populations, though results remain heterogeneous.^28,29 Future research should evaluate the clinical relevance of these activity patterns in larger and more diverse cohorts with extended follow-up to determine long-term outcomes. Exploring behavioral, psychological, and environmental factors that influence activity patterns may further refine patient stratification and inform targeted interventions. Additionally, combining sensor-derived activity data with biomechanical and imaging biomarkers could contribute to developing a multidimensional framework for recovery profiling and advance personalized rehabilitation in orthopedic care.

This study has several limitations that should be considered when interpreting the findings. While the sample size was adequate for primary analyses, it remains modest by physical activity research standards and may limit generalizability, especially for PROMS. Second, the study population was limited to individuals who were independently mobile and capable of using smartphone-based technology, potentially excluding patients with severe functional impairments or limited digital access. Third, the three-month follow-up period, though appropriate for capturing early recovery dynamics, may not reflect longer-term outcomes. Fourth, clinically meaningful thresholds have not yet been established for the wearable-derived activity outcomes used in this study, which limits interpretation of the clinical relevance of the observed between-group differences. Fifth, physical activity classification relied on the vendor-provided SENS Motion processing algorithm. Because full technical details of the classification pipeline were not available for independent inspection, complete algorithmic transparency was limited. Finally, although the clustering approach produced clinically interpretable clusters, it is inherently data-driven and subject to modeling assumptions, which may affect its applicability to other populations or sensor platforms.

5. Conclusion

This study found that patients who were more active before surgery generally showed more favorable objective recovery during follow-up, whereas these differences were not reflected in patient-reported outcomes. This supports the value of combining wearable-derived measures with PROMs to obtain a more complete picture of recovery. As a hypothesis-generating study, these findings draw attention to the potential value of continuous activity monitoring in recovery assessment and support further investigation in larger cohorts.

Footnotes

ORCID iDs

Arash Ghaffari

Andreas Kappel

Shima Gholinezhad

John Rasmussen

Pernille Damborg Clasen

Søren Kold

Ole Rahbek

Ethical considerations

Ethical approval was granted by the North Denmark Regional Committee on Health Research Ethics (Journal number: N-20230035). All study procedures complied with the Declaration of Helsinki and the General Data Protection Regulation. Written informed consent was obtained from all participants prior to enrollment.

Author contributions

All authors conceptualized and designed the study. The methodology was developed by AG, AK, SG and JR. PDC led the data collection. Data analysis was performed by AG and SG. The manuscript was drafted by AG and SG and revised by all co-authors. Supervision was provided by SK, OR, AK, and JR.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.*

Use of AI tools

AI-based language assistance was used for language refinement. All content was reviewed and verified by the authors, who take full responsibility for the manuscript.

References

Price

Alvand

Troelsen

, et al. Knee replacement. Lancet 2018; 392: 1672–1682. https://doi.org/10.1016/S0140-6736(18)32344-4

Shan

Shi

Zhang

. Influencing factors on the quality of recovery after total knee arthroplasty: development of a predictive model. Front Med (Lausanne); 11: 1427768. https://doi.org/10.3389/FMED.2024.1427768

Wainwright

Kehlet

. Functional recovery following hip and knee arthroplasty: subjective vs. objective assessment? Acta Orthop 2022; 93: 739–741. https://doi.org/10.2340/17453674.2022.4567

Judge

Arden

Cooper

, et al. Predictors of outcomes of total knee replacement surgery. Rheumatology (Oxford) 2012; 51: 1804–1813. https://doi.org/10.1093/rheumatology/kes075

Oka

Ono

Tsuboi

, et al. Effect of preoperative sedentary behavior on clinical recovery after total knee arthroplasty: a prospective cohort study. Clin Rheumatol 2020; 39: 891–898. https://doi.org/10.1007/s10067-019-04849-y

Hofstede

Gademan

MGJ

Stijnen

, et al. The influence of preoperative determinants on quality of life, functioning and pain after total knee and hip replacement: a pooled analysis of Dutch cohorts. BMC Musculoskelet Disord 2018; 19: 68. https://doi.org/10.1186/s12891-018-1991-0

Kennedy

Stratford

Hanna

, et al. Modeling early recovery of physical function following hip and knee arthroplasty. BMC Musculoskelet Disord; 7: 100. https://doi.org/10.1186/1471-2474-7-100

Glass

Segal

Callaghan

, et al. Comparison of the extent to which total hip and total knee arthroplasty restore patient-reported physical function. Osteoarthritis Cartilage 2016; 24: 1875–1882. https://doi.org/10.1016/j.joca.2016.06.010

Abeles

Kwasnicki

Pettengell

, et al. The relationship between physical activity and post-operative length of hospital stay: A systematic review. International Journal of Surgery 2017; 44: 295–302. https://doi.org/10.1016/j.ijsu.2017.06.085

10.

Vasta

Papalia

Torre

, et al. The Influence of Preoperative Physical Activity on Postoperative Outcomes of Knee and Hip Arthroplasty Surgery in the Elderly: A Systematic Review. J Clin Med; 9: 969. https://doi.org/10.3390/JCM9040969

11.

Natarajan

Fonseka

Maharaj

, et al. Continuous data capture of gait and mobility metrics using wearable devices for postoperative monitoring in common elective orthopaedic procedures of the hip, knee, and spine: a scoping review. J Orthop Surg Res; 18: 812. https://doi.org/10.1186/S13018-023-04303-5

12.

Lyman

Hidaka

Fields

, et al. Monitoring Patient Recovery After THA or TKA Using Mobile Technology. HSS J 2020; 16: 358–365. https://doi.org/10.1007/s11420-019-09746-3

13.

Zhang

Xie

, et al. Wearable Health Devices in Health Care: Narrative Systematic Review. JMIR Mhealth Uhealth; 8: e18907. https://doi.org/10.2196/18907

14.

Shim

Fleisch

Barata

. Wearable-based accelerometer activity profile as digital biomarker of inflammation, biological age, and mortality using hierarchical clustering analysis in NHANES 2011-2014. Sci Rep 2023; 13: 9326. https://doi.org/10.1038/S41598-023-36062-Y

15.

Wang

Belsky

, et al. Blunted Rest-Activity Circadian Rhythm Is Associated With Increased Rate of Biological Aging: An Analysis of NHANES 2011-2014. J Gerontol A Biol Sci Med Sci 2023; 78: 407–413. https://doi.org/10.1093/gerona/glac199

16.

Hammett

Simonian

Austin

, et al. Changes in Physical Activity After Total Hip or Knee Arthroplasty: A Systematic Review and Meta-Analysis of Six- and Twelve-Month Outcomes. Arthritis Care Res (Hoboken) 2018; 70: 892–901. https://doi.org/10.1002/acr.23415

17.

Ghaffari

Rahbek

Lauritsen

REK

, et al. Criterion Validity of Linear Accelerations Measured with Low-Sampling-Frequency Accelerometers during Overground Walking in Elderly Patients with Knee Osteoarthritis. Sensors 2022; 22(14): 5289. https://doi.org/10.3390/s22145289

18.

Pedersen

Kristensen

Josefsen

, et al. Validation of Two Activity Monitors in Slow and Fast Walking Hospitalized Patients. Rehabil Res Pract 2022; 2022: 9230081. https://doi.org/10.1155/2022/9230081

19.

Hilden

Schwartz

Pascual

, et al. How many days are needed? Measurement reliability of wearable device data to assess physical activity. PLoS One; 18: e0282162. DOI: 10.1371/JOURNAL.PONE.0282162.

20.

Ward

. Hierarchical Grouping to Optimize an Objective Function. J Am Stat Assoc 1963; 58: 236–244. https://doi.org/10.2307/2282967

21.

Harris

Delacqua

Taylor

, et al. The REDCap Mobile Application: a data collection platform for research in regions or situations with internet scarcity. JAMIA Open 2021; 4: ooab078. https://doi.org/10.1093/jamiaopen/ooab078

22.

Boersma

Brouwer

Koolhaas

, et al. No association between preoperative physical activity level and time to return to work in patients after total hip or knee arthroplasty: A prospective cohort study. PLoS One; 14: e0221932. https://doi.org/10.1371/JOURNAL.PONE.0221932

23.

Rooks

Huang

Bierbaum

, et al. Effect of preoperative exercise on measures of functional status in men and women undergoing total hip and knee arthroplasty. Arthritis Care Res (Hoboken): Arthritis Care & Research: Official Journal of the American College of Rheumatology 2006; 55(5): 700–708. https://doi.org/10.1002/art.22223

24.

Onerup

Bock

Börjesson

, et al. Is preoperative physical activity related to post-surgery recovery? a cohort study of colorectal cancer patients. Int J Colorectal Dis 2016; 31: 1131–1140. https://doi.org/10.1007/s00384-016-2551-4

25.

Christensen

Blackburn

Anderson

, et al. Recovery Curve for Patient Reported Outcomes and Objective Physical Activity After Primary Total Knee Arthroplasty—A Multicenter Study Using Wearable Technology. Journal of Arthroplasty 2023; 38: S94–S102. https://doi.org/10.1016/j.arth.2023.03.060

26.

Bin sheeha

Granat

Williams

, et al. Does free-living physical activity improve one-year following total knee arthroplasty in patients with osteoarthritis: A prospective study. Osteoarthr Cartil Open; 2: 100065. https://doi.org/10.1016/j.ocarto.2020.100065.

27.

Harding

Holland

Delany

, et al.

Do activity levels increase after total hip and knee arthroplasty?

Clin Orthop Relat Res 2014; 472: 1502–1511. https://doi.org/10.1007/s11999-013-3427-3

28.

Moyer

Ikert

Long

, et al. The Value of Preoperative Exercise and Education for Patients Undergoing Total Hip and Knee Arthroplasty: A Systematic Review and Meta-Analysis. JBJS Rev 2017; 5: E2. https://doi.org/10.2106/JBJS.RVW.17.00015

29.

Wang

Lee

Zhang

, et al. Does preoperative rehabilitation for patients planning to undergo joint replacement surgery improve outcomes? A systematic review and meta-analysis of randomised controlled trials. BMJ Open; 6: e009857. https://doi.org/10.1136/BMJOPEN-2015-009857