Feasibility and Usability of Wearable Devices for Ambulatory

Monitoring of the Rehabilitation Process of Older Patients after Hip

Fracture Surgery

Dieuwke van Dartel

1,2 a

, Johannes H. Hegeman

1,2

and Miriam M. R. Vollenbroek-Hutten

1,3

Biomedical Signals and Systems Group, University of Twente, Enschede, The Netherlands

Department of Trauma Surgery, Ziekenhuisgroep Twente, Almelo, The Netherlands

ZGT Academy, Ziekenhuisgroep Twente, Almelo, The Netherlands

Keywords: Wearable Devices, Ambulatory Monitoring, Fitbit, MOX, Feasibility, Usability, Older Patients, Hip Fracture.

Abstract: Objective: To assess the feasibility and usability of wearable devices for ambulatory monitoring of older

patients during geriatric rehabilitation after hip fracture surgery.

Methods: Patients (≥70 years) who were surgically treated for a hip fracture wore the Fitbit Charge 2/HR and

the MOX device. Feasibility was assessed by investigating whether real world data gathering revealed

sufficient high-quality data. Usability was assessed by 1) evaluating whether changes in the device parameters

correlated with changes in clinimetric tests and 2) determining whether the wearable devices properly

measured activity.

Results: Data from 67 patients was used to assess feasibility; all patients wore the Fitbit and 33 the MOX.

The mean amount of high-quality data was 88.1% for the Fitbit and 93.6% for the MOX. Data from 42 patients

was used to assess usability; all patients wore the Fitbit and 14 the MOX. A positive progression in clinimetric

tests was correlated with an increase in activity parameters. However, the Fitbit often miscalculated the

number of steps and the MOX algorithm often misclassified slow walking as standing.

Conclusions: Ambulatory monitoring using the Fitbit and MOX is feasible in older patients with a hip fracture.

Concerning the usability, the Fitbit often miscalculated the number of steps. The MOX was more adequate

but the activity classification algorithm often misclassified slow walking based on which it is recommended

to use the raw data instead.

1 INTRODUCTION

The ultimate goal of hip fracture treatment in older

patients is functional recovery, which is defined as the

patient regaining the premorbid level of functioning

(Ceder, 2005; Folbert et al., 2011; Zuckerman, 1996).

To achieve this, adequate post-operative

rehabilitation during and after the patient’s hospital

stay is essential (Prestmo et al., 2015). Clinimetric

tests are often used to obtain insight into the patient’s

progress during rehabilitation. These tests assess the

patient’s physical function, mobility, and cognitive

impairment. Scientific studies have used clinimetric

tests to provide insight into the rehabilitation process

and identify predictive factors for a positive outcome.

However, although clinimetrics provide helpful

https://orcid.org/0000-0002-3556-4522

information, they are also static and administered

infrequently. Furthermore, it is not always possible to

perform a clinimetric test, as patients need a certain

level of mobility (Benzinger et al., 2014;

Hershkovitz, Beloosesky, & Brill, 2012; Nygard,

Matre, & Fevang, 2016). As a result, important

information about patient recovery during

rehabilitation might be missed, with the consequence

that treatment is not adjusted at the right time and

recovery is suboptimal. Therefore, there is a need for

a better, continuous, and accurate way to monitor

older hip fracture patients during rehabilitation.

One possible solution is the use of wearable

devices. Wearable devices are small, portable, body-

fixed sensors that can be used for continuous

ambulatory monitoring of bodily signals. In the case

of hip fracture rehabilitation, most ambulatory

van Dartel, D., Hegeman, J. and Vollenbroek-Hutten, M.

Feasibility and Usability of Wearable Devices for Ambulatory Monitoring of the Rehabilitation Process of Older Patients after Hip Fracture Surgery.

DOI: 10.5220/0010522500590066

In Proceedings of the 18th International Conference on Wireless Networks and Mobile Systems (WINSYS 2021), pages 59-66

ISBN: 978-989-758-529-6

monitoring is performed in the physical activity

domain. Some scientific studies of the rehabilitation

process in older hip fracture patients have already

investigated the added value of physical activity

monitoring. However, these studies monitored

patients for only a few days during hospital stay or

rehabilitation instead of continuously throughout the

whole rehabilitation process, which lasts for several

weeks (Bakker, Blokhuis, Meeks, Hermens, &

Holtslag, 2014; Benzinger et al., 2014; Davenport et

al., 2015; Fleig et al., 2016; Keppler et al., 2020;

Schmal et al., 2018; Talkowski, Lenze, Munin,

Harrison, & Brach, 2009; Taylor, Peiris, Kennedy, &

Shields, 2016). Other studies measured physical

activity in older patients who returned to the

community setting, but again only for a few days

(Fleig et al., 2016; Resnick et al., 2011; Taraldsen et

al., 2015). There is limited information on whether it

is feasible or useful to monitor patients throughout the

entire rehabilitation period using wearable devices.

Armitage et al. recently assessed the feasibility

and acceptability of an activity tracker worn as a

pendant for the continuous monitoring of older

patients (Armitage et al., 2020). In that study, patients

discharged to their home after surgery and patients

temporarily discharged for geriatric rehabilitation

were continuously monitored for 16 weeks. Results

showed that the activity tracker was able to monitor

meaningful activity data. However, many patients

were unwilling to wear it and, therefore, patient

recruitment and retention was low. Therefore, the aim

of this study was to assess the feasibility and usability

of wrist-worn and thigh-worn wearable devices for

the continuous monitoring of older patients during the

entire rehabilitation period after hip fracture surgery.

Feasibility will be assessed by determining whether

real world data gathering revealed sufficient high-

quality data to monitor a patient’s rehabilitation

progression. Usability will be assessed by 1)

evaluating whether changes measured with wearable

devices correlate with changes in the standard

clinimetric tests and 2) by determining whether the

wearable devices properly assess different activities

by comparing recorded data with observations made

during therapy sessions.

2 METHODS

2.1 Subjects

This study included patients aged 70 years or older

who received surgery for their hip fracture at the

department of Trauma Surgery in Ziekenhuisgroep

Twente (ZGT). Patients with severe cognitive

impairment, total hip replacement, a pathological or

periprosthetic fracture, terminal illness, or contact

isolation were excluded. Where possible, patients

were enrolled in the study one day post-surgery; if not

possible, inclusion took place one day before the

patient was discharged for rehabilitation to one of the

three collaborating skilled nursing homes

(TriviumMeulenbeltZorg, Carintreggeland, and

ZorgAccent). If an included patient was not admitted

to one of three collaborating nursing homes,

measurements were only taken during the hospital

stay. This study was part of the “Up&Go after a hip

fracture” project. All patients gave written informed

consent to participate. This study was approved by the

ethical review committee of ZGT and the Medical

Research Ethics Committee Twente.

2.2 Continuous Monitoring

Patients were continuously monitored during their

entire hospital stay and/or rehabilitation stay at the

nursing home. We initially started monitoring

patients with the Fitbit Charge 2 / HR (Fitbit Inc., San

Francisco, CA, USA), which are wrist-worn

community-based activity trackers that were

preferably placed on the patient’s non-dominant

wrist. The Fitbit contains a 3D-accelerometer and

photoplethysmography in order to measure the

number of steps a patient takes and the patient’s heart

rate, respectively. The Fitbit was connected via

Bluetooth to the Fitbit App on a mobile phone to

access the step count and heart rate data.

After a few months of the study, we began

monitoring any newly enrolled patient with a MOX

device in addition to a Fitbit, since the Fitbit is not

able to monitor time spent in different postures. The

MOX (model MMOXX1) is a small, single-unit,

dust- and waterproof device (35x35x10mm) that

continuously monitors physical activity throughout

the day (Maastricht Instruments BV, the

Netherlands). The MOX contains a 3D-

accelerometer, has a sample frequency of 25 Hz, and

was attached to the anterior thigh, 10 cm above the

knee of the fractured leg, with a plaster. We used the

IDEEQ software provided by Maastricht Instrument

BV to download the raw acceleration data from the

MOX and convert it into continuous activity data, i.e.

the number of active minutes (walking) and the

number of sedentary minutes (sitting and lying).

2.3 Assessment of Feasibility

Feasibility was assessed by calculating the amount of

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

high-quality data that was available when older hip

fracture patients were continuously monitored. For

each patient, we first calculated the amount of

missing data for the MOX and Fitbit during daytime

(7.00 am to 10.00 pm) by calculating the number of

missing minutes for each hour. There are no

guidelines in the literature for how to handle missing

data from these devices, so based on our best

judgement, we considered an hour as “missing” if

more than 10 minutes of data were missing. When

more than three hours were missing on a given day,

we considered the day as a missing day. Based on the

number of missing days, we then calculated the

percentage of available data for each patient and the

mean percentage across all patients. The first and last

day of the measurement period were excluded for all

patients because these days were not full

measurement days. Data was analysed with

MATLAB R2017b (MathWorks, Natick, MA, USA).

2.4 Assessment of Usability

Usability was assessed by determining whether the

changes in activity parameters correlated with

changes measured in clinimetric tests, which are

considered the gold standard for evaluating patient

recovery. For this part of the study, we only used data

of the patients monitored during rehabilitation at the

nursing home. For each patient the number of active

minutes, the number of sedentary minutes, and the

number of steps per day were calculated. The

parameters were then used in linear regression, with

time as a dependent variable, to calculate the slope.

The slope was used to determine if the patient’s

progression was positive or negative for each activity

parameter. “Positive progression” was defined as

cases where the number of active minutes and the

number of daily steps have a positive slope and the

number of sedentary minutes a negative slope.

Activity progression was compared to results

from the following clinimetric tests: Timed Up and

Go test (TUG), 10 Meter Walk Test (10MWT),

Functional Ambulation Categories (FAC), Katz

Index of Independence in Activities of Daily Living

(Katz-ADL) and Barthel Index (BI).

The TUG and 10MWT are both functional

capacity tests. For the TUG patients were instructed

to stand up from a chair, walk three meters, turn

around, walk back to the chair, and sit down again.

The time (in seconds) that it took to perform the test

was used as an outcome measure. For the 10MWT the

patient’s gait speed (m/sec) was assessed over a 10-

meter distance and used as an outcome measure.

The FAC, Katz-ADL and BI are functional perfor-

mance tests. The FAC assessed the patient’s ability to

walk and ranged from 0 (not functionally able to

walk) to 5 (walk independently). The TUG and

10MWT tests could only be performed with FAC 3.

The Katz-ADL and BI assessed the patient’s

independence in activities of daily living (ADL). The

Katz-ADL ranged from 0 (completely independent)

to 6 (completely dependent) and the BI from 0

(completely dependent) – 20 (completely

independent).

To calculate patient progression for each

clinimetric test, we calculated the difference between

the test score obtained at discharge from the

rehabilitation department and the test score obtained

at admission to the rehabilitation department.

Differences were expressed as a percentage of the

initial (admission) score, resulting in measurements

for TUG, 10MWT, FAC, Katz-ADL, and BI.

A patient exhibited a “positive progression” during

rehabilitation if 10MWT, FAC, and BI were

positive and TUG and Katz-ADL were negative.

Subsequently, we calculated Pearson’s or Spearman’s

correlation coefficient between the slope of the

activity parameters over time and TUG, 10MWT,

FAC, Katz-ADL, and BI to assess how well

results from continuous sensors correlated with

results from clinimetric tests.

To assess whether the Fitbit and the MOX

correctly identified patient activity as “activity” a

researcher observed weekly therapy sessions at the

rehabilitation department, with 10 patients observed

for a total of 37 sessions. The observer noted the start

and end time for each activity (sitting, standing, and

walking) and manually counted the number of steps

when patients were walking. Results from these direct

observations were compared with the activity-data

logged by the Fitbit and the MOX. Deviations

between the observed and the monitored values were

expressed as percentages.



3 RESULTS

3.1 Subjects

A total of 86 patients were enrolled in this study. Of

these patients, 19 did not complete the study; reasons

for non-completion included problems with

synchronizing the Fitbit (n=6), choosing not to

complete the study (n=3), not wearing the Fitbit

(n=2), discomfort of the Fitbit (n=1), an allergic

reaction to the MOX plaster (n=1), overall decline in

health status (n=1), contact isolation (n=1), death

Feasibility and Usability of Wearable Devices for Ambulatory Monitoring of the Rehabilitation Process of Older Patients after Hip Fracture

Surgery

during rehabilitation (n=1), or unknown reasons

(n=3).

The sensor data from the remaining 67 patients

was used to assess the feasibility of the sensors. All

67 patients wore the Fitbit. The median measurement

period was 24 days (min: 2 days, max: 75 days).

Because the MOX measurements were added later in

the study, only 33 of the 67 patients also wore the

MOX. The median measurement period was 6 days

(min: 2 days, max: 75 days).

Data from 42 of the 67 patients was used to assess

the usability of the wearable sensors. All 42 patients

wore the Fitbit, and 14 of these patients also wore the

MOX. The median measurement period was 29 days

(min: 11 days, max: 71 days) and 27 days (min: 11

days, max 67 days) for the Fitbit and MOX,

respectively. The mean age of the 42 patients was 82

 6 years, and 83% of the patients were female. Prior

to the hip fracture, 69% of the patients lived

independent and 57% of the patients were able to

walk independently. The mean age of the ten patients

whose therapy sessions were observed by a researcher

was 83 ± 3 years, and 70% of the observed patients

were female.

3.2 Feasibility

The percentage of available data varied among

patients, with a maximum of 100% data availability

for both the Fitbit and the MOX and a minimum of

20% data availability for the Fitbit and 74% for the

MOX (Figure 1). The mean percentage of available

data across all patients were 88.1% and 93.6% for the

Fitbit and MOX, respectively. Data availability was

more variable among patients monitored with the

Fitbit compared with the MOX (Figure 1).

Figure 1: The percentage of available data for each patient

for the Fitbit and the MOX.

3.3 Usability

Results show that most patients show a positive

progression throughout their rehabilitation, measured

both with the clinimetric tests as well as with the

activity parameters. However, 10MWT is missing

for 57% of the patients and TUG for 55%.

Figure 2: This figure provides a scatterplot for each activity

parameter, in which the slope of the activity parameter is

compared against the patient’s progression in clinimetric

tests. The clinimetric data is standardized so that the

progression is shown for all the clinimetric tests. The legend

in every quarter shows the percentage of patients within that

plane.

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

Figure 2 presents the scatterplots to compare each

patient’s progression in the clinimetric tests with their

progression in the activity parameters. In each plot,

each point represents a patient’s result. Different

point shapes represent the different clinimetric tests.

Each scatterplot is divided into four quarters; patients

(points) in the green quarters show the same

progression in their activity parameter as in their

clinimetric tests. The pink quarters represent those

patients with discrepancies between activity

parameter and clinimetric tests.

For most patients, the physical activity parameters

show the same progression as the clinimetric tests.

However, approximately 25% of the patients show a

decrease in the number of steps even though the

clinimetric tests indicate a positive progress.

Results from the correlation tests show that for

ΔKatz-ADL there is a moderate negative correlation

with the slope of the number of active minutes (r = -

0.66, p < 0.05, n=14) and a moderate positive

correlation with the slope of the number of sedentary

minutes (r = 0.67, p < 0.01, n=14). For ΔBI it shows

that there is a moderate positive correlation with the

slope of the number of active minutes (r = 0.54, p <

0.05, n=14) and a moderate negative correlation with

the slope of the number of sedentary minutes (r = -

0.57, p < 0.05, n=14). No other significant correlations

were found between the changes in activity parameter

and changes in clinimetric tests (Table 1).

Figure 3: This figure compares the observed activity of

patients during their therapy sessions and the activity

measured by the Fitbit and MOX. For each measure, the

bars show the sum of the activity measure over all observed

therapy sessions.

Figure 3 shows how well the activities are

measured and recorded by the Fitbit and the MOX

compared to direct observations of patient activity.

Data from 24 of the 37 observed therapy sessions

were used to evaluate the accuracy of the Fitbit’s

measured number of steps. Not all therapy sessions

were used since some sessions did not contain proper

step count observations. A total number of 4,202

steps were observed by the researcher across all 24

sessions; however, the Fitbit only counted 942 steps

(Figure 3), which means that only 22.4% of the

observed number of steps were correctly measured by

the Fitbit. The Fitbit generally counted too few steps

when patients were walking with a walker and too

many when patients were moving around in a

wheelchair.

Data from 33 of the 37 observed therapy sessions

were used to compare the activity measured with the

MOX to observed activity. During these sessions, the

researcher observed a total of 6,383 seconds of

sitting, 1,346 seconds of standing, and 7,312 seconds

of walking, whereas the MOX measured 6,533,

6,092, and 2,412 seconds of sitting, standing, and

walking, respectively (Figure 3). This means that the

MOX overestimated the amount of time spent sitting

and standing by 2.3% and 352.6%, respectively, and

underestimated the seconds of walking, as only 33%

of the observed second

s of walking were also

measured by the MOX.

4 DISCUSSION

The aim of this study was to assess the feasibility and

usability of the Fitbit Charge and the MOX for

continuously monitoring of the physical activity of

older patients throughout their rehabilitation period

after hip fracture surgery. We found that 78% of the

patients adhere to the sensors and approximately 88%

and 94% high quality data was available for the Fitbit

and MOX measurements, respectively. This suggests

that it is feasible to use wearable devices for long-

term monitoring and that these devices record enough

data to obtain insight in a patient’s progression during

rehabilitation. We also found that the clinical

progression measured using the sensor parameters

was generally the same as the progression measured

with the standard clinimetric tests, suggesting that the

data produced by the Fitbit and MOX is also usable.

However, the Fitbit was not always able to properly

count the number of steps, especially in patients using

a wheelchair or walking aids, and the IDEEQ

software for analysing the MOX data often classified

slow walking as standing.

4.1 Feasibility

Patients in our study were generally open to wearing

the Fitbit and the MOX sensors and wore them

correctly. Similar results were found by O’Brien et al.

who also showed a high acceptability of a wristband

activity tracker in older adults (O'brien, Troutman-

Feasibility and Usability of Wearable Devices for Ambulatory Monitoring of the Rehabilitation Process of Older Patients after Hip Fracture

Surgery

Table 1: Correlation coefficients from the different correlation tests, which tested whether the slope of the activity parameters

of the Fitbit and the MOX were correlated with the patient’s progression in clinimetric tests. The MOX parameters (slope of

the active minutes and slope of the sedentary minutes) were compared to clinimetric tests using Spearman’s correlation. The

slope of the number of steps was tested using Pearson’s correlation with the exception of the comparison with TUG.

 10MWT



TUG



FAC



KATZ-ADL  BI

Slope number of steps

r=0.02

=0.94

r=0.33

=0.18

r=-0.31

=0.06

r=0.13

=0.42

r=-0.14

=0.39

Slope active minutes

r=-0.2

=0.63

r=0.33

=0.38

r=0.07

=0.83

r=-0.66

p<0.05

r=0.54

p<0.05

Slope sedentary minutes

r=0.39

=0.35

r=-0.08

=0.83

r=-0.28

=0.35

r=0.67

p<0.01

r=-0.57

p<0.05

Jordan, Hathaway, Armstrong, & Moore, 2015). In

contrast, Raymond et al. and Armitage et al. found a

low acceptability of the sensors used in their study.

However, Raymond et al. used an activity tracker that

consists of two parts connected via an electrical cable

(PAL2) and Armitage et al. used an activity tracker

worn in a pendant (Armitage et al., 2020; Raymond,

Winter, Jeffs, Soh, & Holland, 2018). Both devices

were no compact sensor, and both have cables, which

could explain the discrepancy with our generally high

rate of devices acceptance.

We found that wearable devices resulted in more

available data than was obtained using clinimetric

tests. It is not possible to obtain clinimetric data from

every patient since some patients lack the mobility

necessary to perform a clinimetric test. This was true

in our study, as TUG and 10MLT were missing for

55% and 57% of the patients, respectively. However,

this was not the case for continuous monitoring with

the Fitbit and the MOX, for which approximately

88% and 94% of the data were available, respectively.

This shows the advantage of ambulant sensing as this

reveals a high amount of data that do provide a clear

insight in the progression of the patient and enables

the detection of deterioration at an earlier stage.

Data availabilty was not 100% in all patients. The

main reason for missing data was due to sensor

charging; the MOX and Fitbit had a battery life of 7

days and 3-7 days, respectively. The Fitbit exhibited

more variability in the amount of available data for

each patient, which could also be due to forgetting to

synchronize the Fitbit, or due to disturbances in the

measurements caused by sweat, movement of the

sensor, no proper contact with the skin, or excessive

pressure on the skin (Allen, 2007; Jo, Lewis, Directo,

Kim, & Dolezal, 2016).

4.2 Usability

Regaining the premorbid level of functioning is the

main goal in the rehabilitation of an older hip fracture

patient (Ceder, 2005; Zuckerman, 1996), and one way

to achieve this is by increasing physical activity.

Correlation tests showed that an improvement on the

Katz-ADL and BI tests was correlated with a positive

progress in the number of active minutes per day and

the number of sedentary minutes per day. This

corresponds with previous studies that have shown

that physically active patients need less time to regain

ADL function, instrumental ADL function, and

mobility (Hardy & Gill, 2005; Talkowski et al., 2009;

Willems, Visschedijk, Balen, & Achterberg, 2017).

The correlations between the activity parameters

of the MOX and the other clinimetric tests showed the

same directional association, indicating that the

general progression recorded by each approach was

the same, though these correlations were not

significant. This lack of significance can probably be

explained by the fact that the TUG and 10MWT

are focused on the physical capacity of a patient, i.e.

what a patient is capable of doing, whereas the MOX

is focused on the patient’s physical activity, i.e. what

a patient actually does. These are two different

aspects, and it could occur that a patient is showing

less physical activity than he/she is capable of doing,

where pain and low motivation are great barriers for

being physically active during hip fracture

rehabilitation (Benzinger et al., 2014; Resnick et al.,

2011; Sims-Gould, Stott-Eveneshen, Fleig,

McAllister, & Ashe, 2017; Talkowski et al., 2009). In

addition, the correlation coefficients were assessed on

the results of only 14 patients for whom both MOX

and clinimetric data was available, which is a very

limited sample size.

There were more discrepancies between the

progression in the number of steps monitored with the

Fitbit and the progression in the clinimetric tests, and

none of the correlations were significant.

Approximately 25% of the patients showed a positive

progression in their clinimetric tests but a negative

progression in their number of steps per day. One

potential cause of this discrepancy is that the Fitbit

was not able to properly count the number of steps in

older patients, as the Fitbit calculated too many steps

when a patient was in a wheelchair and too few steps

when a patient walked with a walker. This

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems

miscalculation likely arises because the Fitbit is worn

around the wrist. Moving around in a wheelchair

results in movement of the wrists, so the Fitbit

incorrectly counts this movement as steps. Walking

with a walker results in no movement of the wrists, so

the Fitbit does not count any steps. Schmal et al.

similarly found that step counts were less accurate in

patients using mobility aids (Schmal et al., 2018). It

is also possible that patients in our study walked too

slowly for the Fitbit to accurately count their steps

(Schmal et al., 2018); Treacy et al. showed that the

Fitbit produced an inaccurate step count relative to

the observed step count in a group of slow-walking

participants with a mean age of 80 years (Treacy et

al., 2017). We therefore suggest using a wearable

device located on the lower extremities to monitor the

physical activity of older patients.

The MOX is one such device that can be located

on the lower extremity. However, the IDEEQ

software associated with the MOX device was also

unable to properly detect activity in older patients.

The algorithm for activity classification by the

IDEEQ software was designed based on the activity

of two patient populations with a mean age of

54.216.8 and 609.9 years old (Annegarn et al.,

2011), which is significantly younger than the mean

age of the patients monitored in our study (70 years).

This could explain why slow walking was considered

as standing, as the threshold for “walking” was set too

high to correctly classify it in an older population.

However, the IDEEQ software also provides raw

acceleration data, which could be used to design case-

specific activity classification algorithms.

More broadly, this study showed that continuous

monitoring has several advantages over traditional

clinimetric tests. First, continuous monitoring is not

prone to the ceiling effects common in clinimetric

tests. A “ceiling effect” occurs when patients reach a

high or maximal score on a clinimetric test at the

beginning of rehabilitation, leaving little room for

further improvement. A second advantage of

continuous monitoring is the lack of “floor effects,”

which arise when patients are not able to perform a

clinimetric test. In this study, we found floor effects

for the TUG and the 10MWT, which caused a high

percentage of missing data at admission. In contrast,

physical activity could be monitored in all patients,

despite their level of functioning. Third, continuous

monitoring provides more in-depth information about

the progression of a patient at all times and is

therefore less prone to selectively measuring on a bad

or good day.

5 CONCLUSIONS

Continuous physical activity monitoring with the use

of the Fitbit and the MOX was feasible for older hip

fracture patients throughout their rehabilitation

program. Older patients were largely willing to wear

these devices, resulting in a high amount of available

data, and the rehabilitation progression indicated by

continuous monitoring of physical activity was

similar to the progression measured with clinimetric

tests. Continuous monitoring also provided

information about the patient’s progression, including

fluctuations between days and trajectories over time,

that could not be obtained from clinimetric tests.

However, the Fitbit was less usable than the MOX in

a population of older patients. Because the Fitbit is

worn around the wrist, it often could not properly

measure the number of steps in patients who used

mobility aids. The MOX did not have these

disadvantages, though we recommend developing a

new algorithm that uses the raw accelerometer data to

correctly classify the activity of older patients, as the

MOX could sometimes identify slow walking as

standing. Further research is needed to optimize valid

parameter extraction from continuous monitoring

devices worn by patients with very low physical

activity levels, like those recovering from hip surgery.

ACKNOWLEDGEMENTS

The Up&Go after a hip fracture group: M.M.R.

Vollenbroek–Hutten, J.H. Hegeman, E.C. Folbert, S.

Woudsma, C. de Pagter, M.M. Kemerink op

Schiphorst, S. Gommers, T. Oude Weernink, A.J.M.

Harperink, A.H.S. Oude Luttikhuis, N. den Braber, C.

Pierik, A. Malki, and D. van Dartel.

REFERENCES

Allen, J. (2007). Photoplethysmography and its application

in clinical physiological measurement. Physiological

measurement, 28(3), R1.

Annegarn, J., Spruit, M. A., Uszko-Lencer, N. H., Vanbelle,

S., Savelberg, H. H., Schols, A. M., Meijer, K. (2011).

Objective physical activity assessment in patients with

chronic organ failure: a validation study of a new

single-unit activity monitor. Archives of physical

medicine and rehabilitation, 92(11), 1852-1857. e1851.

Armitage, L. C., Chi, Y., Santos, M., Lawson, B. K., Areia,

C., Velardo, C., . . . Farmer, A. J. (2020). Monitoring

activity of hip injury patients (MoHIP): A sub-study of

the World Hip Trauma Evaluation observational cohort

Feasibility and Usability of Wearable Devices for Ambulatory Monitoring of the Rehabilitation Process of Older Patients after Hip Fracture

Surgery

study. Pilot and Feasibility Studies, 6(1).

doi:10.1186/s40814-020-00612-2

Bakker, A., Blokhuis, T. J., Meeks, M. D., Hermens, H. J.,

& Holtslag, H. R. (2014). Dynamic weight loading in

older people with hip fracture. Journal of rehabilitation

medicine, 46(7), 708-711.

Benzinger, P., Lindemann, U., Becker, C., Aminian, K.,

Jamour, M., & Flick, S. (2014). Geriatric rehabilitation

after hip fracture. Zeitschrift für Gerontologie und

Geriatrie, 47(3), 236-242.

Ceder, L. (2005). Predicting the success of rehabilitation

following hip fractures. Disability and rehabilitation,

27(18-19), 1073-1080.

Davenport, S. J., Arnold, M., Hua, C., Schenck, A., Batten,

S., & Taylor, N. F. (2015). Physical activity levels

during acute inpatient admission after hip fracture are

very low. Physiotherapy Research International, 20(3),

174-181.

Fleig, L., McAllister, M. M., Brasher, P., Cook, W. L., Guy,

P., Puyat, J. H., Ashe, M. C. (2016). Sedentary behavior

and physical activity patterns in older adults after hip

fracture: a call to action. Journal of aging and physical

activity, 24(1), 79-84.

Folbert, E., Smit, R., van der Velde, D., Regtuijt, M.,

Klaren, H., & Hegeman, J. (2011). Multidisciplinary

integrated care pathway for elderly patients with hip

fractures: implementation results from Centre for

Geriatric Traumatology, Almelo, The Netherlands.

Nederlands tijdschrift voor geneeskunde, 155(26),

A3197-A3197.

Hardy, S. E., & Gill, T. M. (2005). Factors associated with

recovery of independence among newly disabled older

persons. Archives of Internal Medicine, 165(1), 106-

112.

Hershkovitz, A., Beloosesky, Y., & Brill, S. (2012).

Mobility assessment of hip fracture patients during a

post-acute rehabilitation program. Archives of

Gerontology and Geriatrics, 55(1), 35-41.

Jo, E., Lewis, K., Directo, D., Kim, M. J., & Dolezal, B. A.

(2016). Validation of biofeedback wearables for

photoplethysmographic heart rate tracking. Journal of

sports science & medicine, 15(3), 540.

Keppler, A. M., Holzschuh, J., Pfeufer, D., Neuerburg, C.,

Kammerlander, C., Böcker, W., & Fürmetz, J. (2020).

Postoperative physical activity in orthogeriatric

patients–new insights with continuous monitoring.

Injury.

Nygard, H., Matre, K., & Fevang, J. M. (2016). Evaluation

of Timed Up and Go Test as a tool to measure

postoperative function and prediction of one year

walking ability for patients with hip fracture. Clinical

rehabilitation, 30(5), 472-480.

O'brien, T., Troutman-Jordan, M., Hathaway, D.,

Armstrong, S., & Moore, M. (2015). Acceptability of

wristband activity trackers among community dwelling

older adults. Geriatric Nursing, 36(2), S21-S25.

Prestmo, A., Hagen, G., Sletvold, O., Helbostad, J. L.,

Thingstad, P., Taraldsen, K., . . . Lamb, S. E. (2015).

Comprehensive geriatric care for patients with hip

fractures: a prospective, randomised, controlled trial.

The Lancet, 385(9978), 1623-1633.

Raymond, M. J., Winter, A., Jeffs, K. J., Soh, S.-E., &

Holland, A. E. (2018). Acceptability of physical

activity monitoring in older adults undergoing inpatient

rehabilitation. Aging clinical and experimental

research, 30(8), 1005-1010.

Resnick, B., Galik, E., Boltz, M., Hawkes, W., Shardell, M.,

Orwig, D., & Magaziner, J. (2011). Physical activity in

the post-hip-fracture period. Journal of aging and

physical activity, 19(4), 373-387.

Schmal, H., Holsgaard-Larsen, A., Izadpanah, K., Brønd, J.

C., Madsen, C. F., & Lauritsen, J. (2018). Validation of

Activity Tracking Procedures in Elderly Patients after

Operative Treatment of Proximal Femur Fractures.

Rehabilitation research and practice, 2018.

Sims-Gould, J., Stott-Eveneshen, S., Fleig, L., McAllister,

M., & Ashe, M. C. (2017). Patient perspectives on

engagement in recovery after hip fracture: a qualitative

study. Journal of aging research, 2017.

Talkowski, J. B., Lenze, E. J., Munin, M. C., Harrison, C.,

& Brach, J. S. (2009). Patient participation and physical

activity during rehabilitation and future functional

outcomes in patients after hip fracture. Archives of

physical medicine and rehabilitation, 90(4), 618-622.

Taraldsen, K., Thingstad, P., Sletvold, O., Saltvedt, I.,

Lydersen, S., Granat, M. H., . . . Helbostad, J. L. (2015).

The long-term effect of being treated in a geriatric ward

compared to an orthopaedic ward on six measures of

free-living physical behavior 4 and 12 months after a

hip fracture-a randomised controlled trial. BMC

geriatrics, 15(1), 160.

Taylor, N. F., Peiris, C. L., Kennedy, G., & Shields, N.

(2016). Walking tolerance of patients recovering from

hip fracture: a phase I trial. Disability and

rehabilitation, 38(19), 1900-1908.

Treacy, D., Hassett, L., Schurr, K., Chagpar, S., Paul, S. S.,

& Sherrington, C. (2017). Validity of different activity

monitors to count steps in an inpatient rehabilitation

setting. Physical therapy, 97(5), 581-588.

Willems, E., Visschedijk, J., Balen, R., & Achterberg, W.

(2017). Physical Activity, Physical Function and Fear

of Falling After Hip Fracture. J Orthop Res Physiother,

3, 031.

Zuckerman, J. D. (1996). Hip fracture. New England

journal of medicine, 334(23), 1519-1525.

WINSYS 2021 - 18th International Conference on Wireless Networks and Mobile Systems