Activity Scores of Older Adults based on Inertial Measurement Unit
Data in Everyday Life
Sandra Hellmers
1 a
, Lianying Peng
1
, Sandra Lau
2
, Rebecca Diekmann
1
, Lena Elgert
3
,
J
¨
urgen M. Bauer
2
, Andreas Hein
1
and Sebastian Fudickar
1 b
1
Assistance Systems and Medical Device Technology, Carl von Ossietzky University Oldenburg, 26129 Oldenburg, Germany
2
Center for Geriatric Medicine, University Heidelberg, 69117 Heidelberg, Germany
3
Peter L. Reichertz Institute for Medical Informatics of TU Braunschweig and Hannover Medical School,
TU Braunschweig, 38106 Braunschweig, Germany
Keywords:
Activity Level, Sedentary Time, Inertial Measurement Unit, Healthy Aging, Machine Learning, Functional
Fitness, Functional Decline.
Abstract:
The trend of an ageing population is becoming more and more obvious. Staying healthy in old age is an
important social issue. Thereby, physical activity is essential for the preservation of physical function. We
developed an algorithm for determining the activity level of seniors in everyday life. The proposed algorithm
is based on machine learning activity detection using inertial measurement unit data. A series of activity
scores is obtained by executing the algorithm from data on the type of activity, total activity time and activity
intensity. To evaluate the performance of the proposed algorithm, a study with 251 participants aged above
70 (75.41 ± 3.88) years was conducted and the correlation between individual activity scores and clinical
mobility assessments was determined. Results showed a relation between the Six Minute Walking Test and
the total score in terms of activity level as well as the walk score. Additionally, the MVPA- and walk-score
show a clear trend regarding the frailty status of the participants. Therefore, these scores are indicators of the
physical function and hence validate the utility of the developed algorithm.
1 INTRODUCTION
Healthy ageing is becoming increasingly important,
due to the demographic change and the increasing
number of older people. In general, maintaining the
fitness and independence of older adults to stay at
home as long as possible is of high relevance. Early
detection of functional changes allow interventions
which improve or at least maintain physical function,
which in consequence may reduce fall rates or the
level of dependency (Beswick et al., 2008). Suffi-
cient exercise is of main importance for healthy age-
ing. The results of Morey et al. (Morey et al., 2008)
suggest that physical activity predicts functional sta-
tus and negative changes in activity levels can indicate
a reduction in physical functions. Similarly, accord-
ing to Koroukian et al. (Koroukian et al., 2016), func-
tional limitations may indicate poor health in older
adults.
a
https://orcid.org/0000-0002-1686-6752
b
https://orcid.org/0000-0002-3553-5131
Early changes in functional status can be detected
by clinical assessments. An alternative approach is
the continuous monitoring of physical activity in ev-
eryday life with sensors, e.g. simple pedometers, ac-
celerometers or presence sensors. By using wearable
accelerometers, duration, intensity and type of activ-
ity can be determined via machine learning. In com-
parison to conventional assessments continuous mon-
itoring reduces personnel and time expenditure and
leads to a higher amount of everyday life results as
technical monitoring avoids an examination situation
and the active contribution of the subject.
This paper presents an unobtrusive approach for
determining the activity level of seniors in everyday
life using an inertial measurement unit. An algorithm
was developed which evaluates the different activities
in daily life. In consequence, the use of this method-
ology may result in early-stage interventions to pre-
vent or reduce muscle loss by developing an individ-
ual training plan as soon as changes of the activity
behavior of older people were identified.
Hellmers, S., Peng, L., Lau, S., Diekmann, R., Elgert, L., Bauer, J., Hein, A. and Fudickar, S.
Activity Scores of Older Adults based on Inertial Measurement Unit Data in Everyday Life.
DOI: 10.5220/0009095505790585
In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 579-585
ISBN: 978-989-758-398-8; ISSN: 2184-4305
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All r ights reserved
579
2 STATE OF THE ART
As already mentioned, changes in the physical ac-
tivity of older adults can be a predictor of muscle
loss and a possible loss of mobility. Thereby, phys-
ical activity can be measured both in terms of ac-
tivity behavior and energy consumption. There are
several methods to estimate the activity. For exam-
ple, the recording of physical activity can be carried
out using movement sensors as well as questionnaires
(Ainsworth, 2009). Acceleration sensors were used
in several studies to measure physical activity (Ya-
sunaga et al., 2017; Loyen et al., 2017; van Ballegooi-
jen et al., 2019) and a method to denote generic physi-
cal activity phenotypes from long-term accelerometer
data was presented in (Marschollek, 2013).
However, Copeland et al. have shown, based on a
literature search, that the sedentary time in the case
of surveys compared to the time measured by de-
vices is estimated too low (Copeland et al., 2017).
When recording activity using questionnaires, the
time of physical activity is often overestimated. The
gold standard for determining energy expenditure is
the double-marked water method using isotopes, but
this is expensive and not suitable for everyday use.
Therefore, a technical measurement of activity supe-
riors questionnaires and clinical measurements and is
preferable.
In literature, the activities were often assigned
by an intensity unit based on their share of energy
metabolism. The intensity unit is defined in MET
(metabolic equivalent of task).
Table 1: Activity classification and corresponding MET-
value.
activity MET
sedentary behavior 1.5
mild physical activity 1.5 to 3.0
moderate to strong physical activity 3.0
A MET is defined as the energy conversion dur-
ing quiet sitting, which means an average oxygen
consumption of about 3.5ml · kg
1
· min
1
or 1 kcal ·
kg
1
· h
1
for an adult (Ainsworth et al., 1993). Table
1 shows a basic classification, whereby physical ac-
tivity can be classified according to their intensity in
sedentary behavior, mild physical activity and moder-
ate to vigorous physical activity. For example, For-
tune et al. have developed an algorithm for the classi-
fication of activity levels for persons with rheumatoid
arthritis, in which the activities are classified in class
A (3-5 METs), class B (2-3 METs) and class C (1-
2 METs) (Fortune et al., 2011). Class A activities are
walking, stair climbing and cleaning tasks, class B ac-
tivities are dressing, washing, drying dishes and class
C are reading and writing.
Although it is difficult to accurately characterize
the amount of physical activity which is required to
maintain functional independence, it seems that mod-
erate to higher activity levels are more effective than
mild activities. It can be assumed that there might be
a threshold of at least moderate activity for significant
outcomes (Paterson and Warburton, 2010). Yasunaga
et al. showed that replacing low sedentary behavior
times (e.g., television and desk work) with medium to
high levels of physical activity can contribute to im-
proving physical function in older adults (Yasunaga
et al., 2017). Loyen et. al have found that one-third
of the participants did not meet the physical activity
recommendations of the World Health Organization
(WHO) based on total time in moderate to vigorous-
intensity physical activity (MVPA), while more than
70% did not meet the recommendations based on
time in MVPA bouts of at least 10 minutes (Loyen
et al., 2017). In most articles with sensor-based ac-
tivity estimation, the time in sedentary behavior (SB),
light-intensity physical activity (LIPA) and moderate
to vigorous-intensity physical activity (MVPA) were
distinguished. Whereby, the wearing period of the
accelerometer was at least four days (including one
weekend day) with at least 10 hours recording (600
minutes) per day.
In most cases, the questionnaires evaluate the fre-
quency, duration and type of activity in order to de-
termine the level of physical activity. In the work
of Huy and Schneider, a tool for measuring activ-
ity was developed for the German-speaking countries
(Huy and Schneider, 2008) and compared following
questionnaires: Dijon Physical Activity Score, Mod-
ified Baecke Questionnaire for Older Adults (MBQ),
Physical Activity Scale for the Elderly (PASE), Yale
Physical Activity Survey (YPAS) and Zutphen Phys-
ical Activity Questionnaire (ZPAQ). Trampisch et
al. compared the MBQ, ZPAQ, PASE, YPAQ and
CHAMPS Physical Activity Questionnaire for Older
Adults (Trampisch et al., 2011). Since the question-
naires MBQ (Voorrips et al., 1991), PASE (Washburn
et al., 1993), YPAS (Dipietro et al., 1993) and ZPAQ
(Caspersen et al., 1991) have been used in both ar-
ticles and are suitable for our target group of older
adults starting at 60 years, they will also be consid-
ered in the present analysis.
These questionnaires identified the following ar-
eas of physical activity: sports, leisure, household,
gardening, occupation, locomotion and rest periods
(Huy and Schneider, 2008). All questionnaires use an
overall score that summarizes all surveyed physical
activities into a single result. Each activity is assigned
HEALTHINF 2020 - 13th International Conference on Health Informatics
580
a weighting. For YPAS and ZPAQ, this weighting is
based on the MET value, which corresponds to the
energy consumption during the performance of the
physical activity and an intensity score was assigned
to each specific activity area. MBQ and PASE re-
sult in a total score. MBQ calculates the sum of the
activity areas sports, household and leisure with dif-
ferent weightings and PASE calculates the sum of 12
activity-subareas with different weightings.
3 METHODS
The aim of the present analysis was the calculation
of an individual activity score based on the performed
activities in daily life. The score should hold crucial
information about the functional status and should be
sensitive for changes in the activity behavior of the
older adults to recognize first signs of functional de-
terioration occurring with age.
The study design and the algorithm for the score
calculation are presented in the following subsections.
3.1 Study Design and Sample
We conducted a study with 251 participants aged 70
at minimum. The characteristics of our study sample
are listed in Table 2. 148 participants were female and
103 male.
Table 2: Characteristics of the study sample (n=251), SD
stands for standard deviation.
min max mean (SD)
age [years] 70 87 75.41 (3.88)
height [cm] 145.8 192.5 167.10 (9.27)
weight [kg] 46.85 123.45 76.85 (14.15)
The participants performed first a comprehensive
clinical assessment, which included besides other
tests the Six Minute Walking Test (6MWT) (Butland
et al., 1982), the Short Physical Performance Battery
(SPPB) (Guralnik et al., 1994), the Timed Up and
Go Test (TUG) (Podsiadlo and Richardson, 1991),
the Frailty Index (FI) (Fried et al., 2001) and the
Stair Climb Power Test (SCPT) (Bean et al., 2007).
All functional tests were measured by physical thera-
pists in a conventional manner as well as technology-
based. The description of the whole study can be
found in (Hellmers et al., 2017). Besides medical
products and ambient sensors like light barriers, an
inertial measurement unit (IMU) integrated into a belt
was also used. The used sensor unit consists of a
3D accelerometer, gyroscope, magnetometer and a
barometer.
Following the assessment, the participants wore
the IMU-based sensor belt for seven days in their ev-
eryday life. To compare the results of the sensor with
a gold standard measure, they were asked to write an
activity diary.
In this analyses, data of 231 participants were in-
cluded. 20 subjects were excluded, due to too short
measurements with a duration below 4 days or a mean
duration of fewer than 10 hours per day, incorrect
wearing of the belt (deviating sensor orientation) or
technical problems. The criterion of a minimum mea-
surement time was chosen on the basis of the litera-
ture research (see Section 2).
3.2 Hierarchical Classification Model
and Score Algorithm
To develop an adequate algorithm for the score cal-
culation, the sensor measurements were analyzed via
a machine learning classifier regarding the performed
activities. We used a hierarchical classification model
with four classifiers. After a low pass filtering of
the raw data, the first classifier distinguished between
static and dynamic activities, as well as transitions.
If the state was classified, the other classifiers char-
acterized the activities in detail. The selection of the
trained activities was chosen regarding the activities
in the conducted assessment tests. Figure 1 illus-
trates the hierarchical classification model. The state
classifier is based on boosted decision trees, whereby
the other classifiers are based on multilayer percep-
trons (MLP). The MLP-classifier for the static activ-
ities consists of 5 hidden layers (HL) and 7 hidden
nodes (HN), the classifier for dynamic activities of 3
HL and 44 HN, and for transitions of 4 HL and 40
HN. Further details of the different classifiers, used
window sizes, step widths, filters and feature sets can
be found in (Hellmers et al., 2018; Hellmers et al.,
2019).
Figure 1: Hierarchical classification model. The first clas-
sifier distinguished between the state. The following classi-
fiers recognized the possible activities of each state.
After the classification of the performed activi-
ties, an activity score can be calculated. Therefore,
each activity is assigned an intensity weighting. The
weightings are based on the MET-values found in the
literature (see Section 2) and listed in Table 3.
Activity Scores of Older Adults based on Inertial Measurement Unit Data in Everyday Life
581
Figure 2: Average time in minutes of each activities per day over 231 participants.
The total activity score S is calculated by the
whole duration of an activity A
i
per day
A
i
=
m
k=1
Z
k
/T, (1)
with Z
k
as duration of activity and T number of mea-
suring days, multiplied by the intensity weighting I
i
S =
n
i=1
(I
i
· A
i
), (2)
with n as total number of recognized activities. Al-
ternatively, specific scores like the sedentary behavior
score can be calculated by only taking the correspond-
ing activities into account:
S
SB
= I
sitting
· A
sitting
+ I
lying
· A
lying
(3)
LIPA-, MVPA-Scores and scores of specific activities
can be calculated analogously.
4 RESULTS
Figure 2 shows the activities per day in minutes over
the total study sample. As expected, the participants
showed most of the time a sedentary behavior with in
average 484.43 ± 126.88 minutes. The second most
frequent activity was standing with 250.30 ± 25.61
minutes per day. The average walking time is about
75.28 ± 25.61 minutes per day. Due to the limited se-
lection of recognized activities, there are some miss-
classifications identified from the diaries of the par-
ticipants. For example, housework and gardening are
recognized as standing in most cases.
Table 3: Intensity weightings and categories of each ac-
tivity. The abbreviations stand for: Sedentary behavior
(SB), light-intensity physical activity (LIPA) and moderate
to vigorous-intensity physical activity (MVPA).
activity intensity category
weighting
sitting 1 SB
lying 1 SB
standing 2 LIPA
walking 3.5 MVPA
transitions 5.5 MVPA
climbing stairs 6.85 MVPA
jumping 10 MVPA
Another important activity for our target group is
cycling. Cycling is often classified as climbing stairs,
walking, sitting or standing. In a self-experiment it
could be shown that the speed of cycling has a sig-
nificant influence on the classification. Thus, very
slow riding is classified as sitting while fast rid-
ing is recognized as climbing stairs. Since driving
faster means a higher energy consumption, this ”miss-
classification” is categorized correctly regarding the
intensity weighting.
In terms of METs, the activities also fall into
the right category: for example, slow-cycling (5.5
mph) corresponds to 3.5 METs, while faster cycling
(10-11.9 mph) corresponds to 6.8 METs (Ainsworth
et al., 2011). These values are similar to the intensity
weightings of 3.5 for walking and 6.85 for climbing
stairs.
In a second analysis, the study participants were
categorized into four different groups regarding the
HEALTHINF 2020 - 13th International Conference on Health Informatics
582
Figure 3: Average MVPA-score, walk-score, stand-score and sedentary behavior-score (SB) for the different frailty groups
regarding the Fried phenotype.
Fried phenotype: robust (0 points) , pre-frail (1 point
and 2 points) and frail ( 3 points). The distribution
of the 231 subjects among the groups is as follows:
160 (robust - 0 points), 60 (pre-frail - 1 point), 11
(pre-frail - 2 points). Figure 3 shows the results of the
MVPA-score, walk-score, stand-score and SB-score
for the different frailty groups. There is a clear trend
that seniors with a low MVPA-score (decreasing ac-
tive) are frailer. On average, the participants of the
robust-group have a higher MVPA-score of 4.32 and
(pre-) frail seniors have lower MVPA-scores of 3.74
and 2.88. This trend can also be observed in the walk-
ing score. Therefore, the walking-score could also be
a predictor of frailty. However, the SB-score seems to
be not a sure sign of frailty. Due to the disbalanced
group sizes, these results should be considered with
caution.
The scores also reflect the long times spent stand-
ing or in a sedentary behavior.
Figure 4 presents the results of the walking score
in comparison to the results of the Six Minute Walk-
ing Test (6MWT). Correlation analysis shows a light
significant correlation with a p-value of p < 0.001
and a correlation coefficient of 0.48. The dashed line
shows the calculated regression line.
Figure 4: Correlation of the walking-score with the results
of the Six Minute Walking Test.
5 CONCLUSIONS
We developed an algorithm for the estimation of an
activity level of older adults based on inertial mea-
surement unit measurements in daily life. We con-
ducted a study with 251 participants aged 70 years
and above and analyzed the activities recognized via
a hierarchical machine learning classification model.
Different activity scores were calculated, whereby the
MVPA- and walk-score show a clear trend regarding
the frailty status of the participants. Correlation anal-
yses with the results of clinical mobility assessments
showed a significant correlation between the walk-
Activity Scores of Older Adults based on Inertial Measurement Unit Data in Everyday Life
583
score and the Six Minute Walking Test.
Therefore, we presented a promising approach
that provides information on the health status through
unobtrusive everyday measurements. Further investi-
gation about the significance of the scores are neces-
sary. Concerning the miss-classifications, an expan-
sion of the recognizable activities would be valuable
to improve the meaningfulness of the scores.
However, we are optimistic that this approach will
be able to detect changes in the activity behaviour of
seniors and thus is suitable to give early indications of
the necessity of interventions in the event of a begin-
ning functional decline.
ACKNOWLEDGEMENTS
The study is funded by the German Federal
Ministry of Education and Research (Project No.
01EL1822D).
REFERENCES
Ainsworth, B. E. (2009). How do i measure physical activ-
ity in my patients? questionnaires and objective meth-
ods. British Journal of Sports Medicine, 43(1):6–9.
Ainsworth, B. E., Haskell, W. L., Herrmann, S. D., Meckes,
N., Bassett Jr, D. R., Tudor-Locke, C., Greer, J. L.,
Vezina, J., Whitt-Glover, M. C., and Leon, A. S.
(2011). 2011 compendium of physical activities: a
second update of codes and met values. Medicine &
science in sports & exercise, 43(8):1575–1581.
Ainsworth, B. E., Haskell, W. L., Leon, A. S., Jacobs, J. D.,
Montoye, H. J., Sallis, J. F., and Paffenbarger, J. R.
(1993). Compendium of physical activities: clas-
sification of energy costs of human physical activi-
ties. Medicine and science in sports and exercise,
25(1):71–80.
Bean, J. F., Kiely, D. K., LaRose, S., Alian, J., and Fron-
tera, W. R. (2007). Is stair climb power a clinically
relevant measure of leg power impairments in at-risk
older adults? Archives of physical medicine and reha-
bilitation, 88(5):604–609.
Beswick, A. D., Rees, K., Dieppe, P., Ayis, S., Gooberman-
Hill, R., Horwood, J., and Ebrahim, S. (2008). Com-
plex interventions to improve physical function and
maintain independent living in elderly people: a
systematic review and meta-analysis. The Lancet,
371(9614):725–735.
Butland, R., Pang, J., Gross, E., Woodcock, A., and Geddes,
D. (1982). Two-, six-, and 12-minute walking tests in
respiratory disease. British medical journal (Clinical
research ed.), 284(6329):1607.
Caspersen, C. J., Bloemberg, B. P., Saris, W. H., Merritt,
R. K., and Kromhout, D. (1991). The prevalence of se-
lected physical activities and their relation with coro-
nary heart disease risk factors in elderly men: the zut-
phen study, 1985. American journal of Epidemiology,
133(11):1078–1092.
Copeland, J. L., Ashe, M. C., Biddle, S. J., Brown, W. J.,
Buman, M. P., Chastin, S., Gardiner, P. A., Inoue, S.,
Jefferis, B. J., Oka, K., et al. (2017). Sedentary time
in older adults: a critical review of measurement, as-
sociations with health, and interventions. Br J Sports
Med, 51(21):1539–1539.
Dipietro, L., Caspersen, C. J., Ostfeld, A. M., and Nadel,
E. R. (1993). A survey for assessing physical activity
among older adults. Medicine & Science in Sports &
Exercise.
Fortune, E., Tierney, M., Scanaill, C. N., Bourke, A.,
Kennedy, N., and Nelson, J. (2011). Activity level
classification algorithm using shimmer
TM
wearable
sensors for individuals with rheumatoid arthritis. In
2011 Annual International Conference of the IEEE
Engineering in Medicine and Biology Society, pages
3059–3062. IEEE.
Fried, L. P., Tangen, C. M., Walston, J., Newman, A. B.,
Hirsch, C., Gottdiener, J., Seeman, T., Tracy, R., Kop,
W. J., Burke, G., et al. (2001). Frailty in older adults:
evidence for a phenotype. The Journals of Gerontol-
ogy Series A: Biological Sciences and Medical Sci-
ences, 56(3):M146–M157.
Guralnik, J. M., Simonsick, E. M., Ferrucci, L., Glynn,
R. J., Berkman, L. F., Blazer, D. G., Scherr, P. A.,
and Wallace, R. B. (1994). A short physical perfor-
mance battery assessing lower extremity function: as-
sociation with self-reported disability and prediction
of mortality and nursing home admission. Journal of
gerontology, 49(2):M85–M94.
Hellmers, S., Fudickar, S., B
¨
use, C., Dasenbrock, L.,
Heinks, A., Bauer, J. M., and Hein, A. (2017). Tech-
nology supported geriatric assessment. In Ambient As-
sisted Living, pages 85–100. Springer.
Hellmers, S., Fudickar, S., Lau, S., Elgert, L., Diekmann,
R., Bauer, J. M., and Hein, A. (2019). Measurement
of the chair rise performance of older people based on
force plates and imus. Sensors, 19(6):1370.
Hellmers, S., Izadpanah, B., Dasenbrock, L., Diekmann, R.,
Bauer, J., Hein, A., and Fudickar, S. (2018). Towards
an automated unsupervised mobility assessment for
older people based on inertial tug measurements. Sen-
sors, 18(10):3310.
Huy, C. and Schneider, S. (2008). Instrument f
¨
ur die er-
fassung der physischen aktivit
¨
at bei personen im mit-
tleren und h
¨
oheren erwachsenenalter. Zeitschrift f
¨
ur
Gerontologie und Geriatrie, 41(3):208–216.
Koroukian, S. M., Schiltz, N., Warner, D. F., Sun, J.,
Bakaki, P. M., Smyth, K. A., Stange, K. C., and Given,
C. W. (2016). Combinations of chronic conditions,
functional limitations, and geriatric syndromes that
predict health outcomes. Journal of general internal
medicine, 31(6):630–637.
Loyen, A., Clarke-Cornwell, A. M., Anderssen, S. A.,
Hagstr
¨
omer, M., Sardinha, L. B., Sundquist, K.,
Ekelund, U., Steene-Johannessen, J., Baptista, F.,
Hansen, B. H., et al. (2017). Sedentary time and
physical activity surveillance through accelerometer
HEALTHINF 2020 - 13th International Conference on Health Informatics
584
pooling in four european countries. Sports Medicine,
47(7):1421–1435.
Marschollek, M. (2013). A semi-quantitative method to de-
note generic physical activity phenotypes from long-
term accelerometer data–the atlas index. PLoS One,
8(5):e63522.
Morey, M. C., Sloane, R., Pieper, C. F., Peterson, M. J.,
Pearson, M. P., Ekelund, C. C., Crowley, G. M.,
Demark-Wahnefried, W., Snyder, D. C., Clipp, E. C.,
et al. (2008). Effect of physical activity guidelines on
physical function in older adults. Journal of the Amer-
ican Geriatrics Society, 56(10):1873–1878.
Paterson, D. H. and Warburton, D. E. (2010). Physical
activity and functional limitations in older adults: a
systematic review related to canada’s physical activity
guidelines. International Journal of Behavioral Nu-
trition and Physical Activity, 7(1):38.
Podsiadlo, D. and Richardson, S. (1991). The timed “up &
go”: a test of basic functional mobility for frail elderly
persons. Journal of the American geriatrics Society,
39(2):142–148.
Trampisch, U. S., Platen, P., Moschny, A., and Hinrichs,
T. (2011). Die eignung von frageb
¨
ogen zur erfassung
der k
¨
orperlichen aktivit
¨
at
¨
alterer erwachsener f
¨
ur den
einsatz in einer epidemiologischen studie. Deutsche
Zeitschrift f
¨
ur Sportmedizin, 62(10):329–333.
van Ballegooijen, A. J., van der Ploeg, H. P., and Visser,
M. (2019). Daily sedentary time and physical activity
as assessed by accelerometry and their correlates in
older adults. European Review of Aging and Physical
Activity, 16(1):3.
Voorrips, L. E., Ravelli, A., Petra, C., Dongelmans, A.,
Deurenberg, P., and van Staveren, W. A. (1991). A
physical activity questionnaire for the elderly. Diet
and physical activity as determinants of nutritional
status in elderly women, page 43.
Washburn, R. A., Smith, K. W., Jette, A. M., and Janney,
C. A. (1993). The physical activity scale for the el-
derly (pase): development and evaluation. Journal of
clinical epidemiology, 46(2):153–162.
Yasunaga, A., Shibata, A., Ishii, K., Koohsari, M. J., Inoue,
S., Sugiyama, T., Owen, N., and Oka, K. (2017). As-
sociations of sedentary behavior and physical activity
with older adults’ physical function: an isotemporal
substitution approach. BMC geriatrics, 17(1):280.
Activity Scores of Older Adults based on Inertial Measurement Unit Data in Everyday Life
585