1

Introduction

Physical activity is an important determinant of health and well-being and inactivity is strongly associated with morbidity and mortality . Inactivity was identified by the World Health Organization as the fourth strongest risk factor for mortality. Several diseases reduce physical activity, osteoarthritis being one of the leading causes and affecting 9% of the population . Restoration of function and activity in patients with these diseases is one of the main goals for clinicians and is specific for patients with osteoarthritis after total joint arthroplasty (TJA). Monitoring physical activity before, shortly after and in late follow-up after TJA is crucial for evaluating rehabilitation, assessing social participation and determining the effectiveness of the interventions in evidenced-based practice.

Self-reported questionnaires are widely used to assess physical activity. Although they are inexpensive and easy to use, they show limited validity and reliability and are subject to bias; they largely overestimate activity . In contrast, the gold standards in assessing physical activity in terms of energy expenditure – ingestion of double labeled water, room calorimetry and indirect calorimetry – are time-consuming, costly and technically too demanding for routine clinical use. Furthermore, these techniques do not provide specific information on lower-limb activity or type of activity such as walking, stair climbing or cycling.

Portable sensors have evolved considerably over the years, from simple step counters to accelerometer-based activity monitors (AMs) combined with heart rate monitors, inertia-based monitors and GPS trackers . Most of these devices are designed to give an estimation of energy expenditure or activity counts . Although they provide data on general physical activity levels, they do not provide information about specific lower-limb motor activities such as walking, transfers from sit to stand or climbing stairs. Monitoring these tasks is fundamental for assessing physical activity in patients with functional limitations in lower extremities.

A number of available AMs can identify specific activities such as sitting, lying down, standing and walking, with reasonably good accuracy , even in free-living settings . Accelerometer-based activity monitoring in impaired patients or shortly after TJA poses an additional challenge. Because these patients move slowly and are crutch-supported, the magnitude of acceleration signals produced by activities will be lower and more difficult to detect. Several validation studies of slow-walking healthy people and patients with chronic obstructive pulmonary disease , older adults with walking aids and patients with rheumatoid arthritis showed poor validity of such monitoring in detecting the number of steps or transitions. Recently, good results have been achieved in slow-walking but only healthy people . Moreover, physical activity of patients after TJA has been monitored during only late follow-up . Therefore, monitoring physical activity in impaired patients requires the validation of a new AM.

The aim of this study was to validate a novel single-sided AM in impaired, slow-walking, crutched-supported patients after TJA.

2

Materials and methods

To validate the AM in impaired and crutched-supported walking, we included patients who gave consent shortly after TJA, a median of 3 to 17 days (mean: 8.9 ± 5.3) after arthroscopy of the knee ( n = 23) or hip ( n = 17). Patients were able to safely move with 2 crutches. Eight healthy people without locomotor impairments were selected as a reference group and resembled fully recovered patients. The groups were not compared, so they were not matched. The study protocol was approved by the medical ethical committee of the Zuyderland medical center and all participants gave informed consent.

Patients and healthy people completed a trial in which they had to perform several activities consisting of walking, standing, resting (sitting and lying supine) and stair climbing (ascending and descending). Participants were free to complete these tasks in a random order, in their own way and at their preferred speed. Furthermore, the duration and number of repetitions of the activities depended on the person’s condition. All activities were performed at least once. The trial was recorded on video, which was used for retrospective analysis. At the start and end of the trial, the AM and video recordings were synchronized by tapping on the sensor.

A commercially available, light-weight (18 g, 64 × 25 × 13 mm) 3D accelerometer (USB accelerometer X16-mini, Gulf Coast Data Concepts, Waveland, MS, USA; Fig. 1 a ) was taped to the lateral non-affected upper leg of participants by using hypo-allergenic double-sided tape (3 M, product number 9917, Fig. 1 b). The position of the sensor was at the mid-thigh between the trochanter and lateral condyle, with the y-axis of the sensor along the axis of the femur. We collected 12-bit data (range ± 2 g) at a sampling rate of 40 Hz; data were stored on an on-board memory micro-SD card. The raw acceleration signal was analyzed by using the inclinometer function of the accelerometer and algorithm-based peak detection methods in MATLAB as described . Briefly, the accelerometer’s orientation is calibrated during a period of level walking that is manually selected. Within this walking period, the average magnitudes of the 3 acceleration vectors and the gait cycle frequency (GCF) are derived to allow for further differentiation between activities. Differentiation between standing periods and resting periods is based on the direction of the gravitation vector and allows for identifying sit-to-stand transfers. Walking is differentiated from other upright activities (all classified as standing) by applying heuristic rules for the GCF. A walking period is classified with the detection of at least 5 consecutive heel strike peaks, with < 1.5 s between peaks for GCF > 0.6 Hz and < 3.0 s between peaks for GCF < 0.6 Hz. Furthermore, ascending and descending stairs is distinguished from level of walking based on the inclinometer function and the hip angle. An example of the raw acceleration signal is shown in Fig. 2 .

The 3D accelerometer (USB accelerometer X16-mini, Gulf Coast Data Concepts, Waveland, MS, USA).

Raw acceleration signals for activities.

Data analysis involved the type and duration of activities (walking, standing, resting, ascending and descending stairs) in seconds, the total number of activities including gait cycles (GCs), steps ascending and descending stairs, walking bouts (level walking, ascending and descending stairs) and transfers (from sitting or lying supine to straight). Activity intensity was assessed by walking cadence (GC × 2/min, determined over walking bouts of ≥ 10 GCs).

The video recordings were observed by one investigator. Activities were classified (type and duration) with a 1-s resolution, and GCs in level walking, steps descending and ascending stairs, bouts (level walking, ascending and descending stairs) and transfers were counted. In a pilot study of 20 patients observed by 2 investigators, inter-rater agreement in classifying activities with video observations was high (interclass correlation coefficients: 99.7–100%), which was comparable to previous studies .

Validity of the AM to identify different activities was assessed by determining sensitivity (the proportion of seconds of correctly classified activities in terms of actual performed activities, as defined by the video observations) and predictive value (the proportion of seconds of correctly classified activities in terms of determined activities by the AM). Because of the impaired function of patients after TJA, transition phases between activities (e.g., from sitting to walking) represent a substantial part of the actual executed activity. In addition, because of these transition phases, the observer cannot accurately determine the start and end of the activities. Therefore, transitional phases (10 s for patients, 4 s for healthy participants) were discarded for the analysis as previously described .

The agreement between the AM and video recordings of the total time of activities (without discarded transitional phases) and the cumulative outcome parameters (total number of sit-to-stand transfers, total GCs, total steps ascending and descending stairs and total number of bouts) was assessed by the mean percentage difference (MPD) with Bland-Altman plots . In addition, we plotted the MPD against mean values. Differences in validity between patients and healthy participants were calculated by ANOVA or Student t test. P < 0.05 was considered statistically significant.

3

Results

We included 40 patients (mean age: 65 ± 9 years; mean BMI: 30 ± 6 kg/m ² ; male:female ratio: 18:22; mean days postoperation: 8.9 ± 5.3) and 8 healthy participants (mean age: 49 ± 20 years; mean BMI: 23 ± 0.7 kg/m ² ; male:female ratio: 5:3). Video observations showed a mean total time of recording of 16.1 ± 5.4 min for patients and 33.2 ± 9.1 min for healthy participants. Patients spent a mean of 5.7 ± 2.5 min resting, 5.5 ± 1.7 min standing and 5.0 ± 2.3 min walking (level walking, ascending and descending stairs) in a mean of 3.8 ± 1.3 level walking bouts, with a mean cadence of 56.0 ± 10.0 steps/min ( Fig. 2 ). A mean of 107.5 ± 46.0 GCs were taken and 2.6 ± 0.7 transfers performed. All patients ascended and descended the stairs once. Healthy participants spent a mean of 10.6 ± 4.0 min resting, 8.7 ± 3.2 min standing, and 9.0 ± 2.1 min walking (level walking, ascending and descending stairs) in a mean of 30.4 ± 7.2 level walking bouts and 8.8 ± 2.4 ascending and 8.6 ± 2.3 descending stair climbing bouts. Healthy participants walked a mean of 344.6 ± 82.0 GCs, with a cadence of 97.9 ± 8.9 steps/min and performed 87.5 ± 23.8 and 86.25 ± 22.6 steps ascending and descending stairs, respectively. In addition, they performed a mean of 8.6 ± 2.3 transfers.

For patients, the mean sensitivity and predictive values were high (> 97%) for all categories ( Table 1 ). In one patient, the sensitivity of descending stairs was moderate (73.7%). All descending stair-climbing bouts were correctly identified by the AM and only one transfer and one ascending stair-climbing bout was not detected. The overall agreement between AM and video recordings in detecting the number of GCs was excellent (MPD = 2.0 ± 4.3%). The AM overestimated the number of level walking bouts (MPD = 2.1 ± 7.3%), number of steps ascending and descending stairs (MPD < 7%) and cadence (MPD = 10.8 ± 5.3%) ( Table 2 ).

Table 1

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

Complex regional pain syndrome associated with hyperattention rather than neglect for the healthy side: A comprehensive case study Care-related pain and discomfort in children with motor disabilities in rehabilitation centres Pelvic floor muscle reflex activity during coughing – an exploratory and reliability study An update on the pathophysiology of osteoarthritis Does the physiatrist replicate the functions of the neurologist or orthopedist? Takotsubo cardiomyopathy as a reversible complication of intrathecal baclofen withdrawal

Stay updated, free articles. Join our Telegram channel

Join