Breaking up Long Sedentary Periods of Office Workers through a
Virtual Coach using Activity Data
Jasmijn Franke
1,2
, Christiane Gr
¨
unloh
1,2 a
, Dennis Hofs
1 b
, Boris Van Schooten
1 c
,
Andreea Bondrea
1
and Miriam Cabrita
1,2,3 d
1
eHealth Group, Roessingh Research and Development, Enschede, The Netherlands
2
Biomedical Signals and Systems Group, University of Twente, Enschede, The Netherlands
3
Innovation Sprint Sprl, Brussels, Belgium
Keywords:
Physical Activity, Sedentary Periods, Sedentary Bouts, Inactive Periods, Inactive Bouts, Virtual Coach,
Coaching, Embodied Conversational Agent, mHealth, eHealth.
Abstract:
Office workers often lead sedentary lifestyles, a lifestyle responsible for higher risks of cardiovascular disease,
stroke, diabetes and premature mortality. Improvements towards a more active lifestyle reduce cardiovascular
risks and thus changing the sedentary lifestyle might prevent chronic illness. The Recurring Sedentary Period
Detection (RSPD) algorithm described in this paper was designed to identify recurring sedentary periods
using data from an activity tracker, summarise the sedentary periods and pinpoint notification times at which
the user should be motivated to get some movement. The outcome of the RSPD algorithm was validated
using data from a 10-week period of one typical office worker. Our results show that the RSPD algorithm
could correctly identify the recurring sedentary periods, compute fitting daily summaries and pinpoint the
notification times correctly. With minor differences, the RSPD algorithm was successfully implemented in the
healthyMe smartphone application, one of the supporting services of the SMARTWORK project. Within the
healthyMe application, an embodied virtual agent is used to communicate the daily summaries and motivate
the user to move more at the identified notification times. Pilots planned as part of the SMARTWORK project
will evaluate whether the RSPD algorithm helps to motivate office workers to break up sedentary periods.
1 INTRODUCTION
An active lifestyle contributes positively to both
mental (Rohrer et al., 2005) and physical health
(Gonz
´
alez-Gross and Mel
´
endez, 2013). Office
workers however, lead a sedentary lifestyle due to
prolonged hours of sitting behind a desk, which may
lead to significantly higher risks of cardiovascular
disease, stroke, diabetes and premature mortality
(Parry and Straker, 2013). Especially working people
above the age of 55 years old have higher risks
of becoming sick or chronically ill (Rogers and
Wiatrowksi, 2005). The systematic literature review
by Barbaresko et al. showed that healthy lifestyle
behaviours (including being physically active) are
associated with a reduced risk of cardiovascular
a
https://orcid.org/0000-0003-2319-3186
b
https://orcid.org/0000-0003-1165-3106
c
https://orcid.org/0000-0001-8131-6637
d
https://orcid.org/0000-0001-6757-9406
diseases (Barbaresko et al., 2018). Thus, certain
chronic illnesses may be prevented by changing the
sedentary lifestyle of office workers. Currently,
there are many available commercial activity trackers
connected to smartphone applications, that give
insight in the user’s lifestyle. However, many of
these applications are not user-friendly for the older
workers with low digital literacy (Kocsis et al., 2019),
are not personalised and their focus is on tracking
and visualising activity, rather than coaching. For
example, many activity tracking applications have a
default daily step-goal of 10,000 steps. Such a goal
can be daunting and de-motivating for people with a
sedentary lifestyle, as they feel they will never be able
to reach it. Thus, a personalised goal that is within
reach is necessary to help the user to reach their daily
goal.
Setting an appropriate goal is only the first step
into changing a person’s sedentary lifestyle (Abraham
and Michie, 2008), the next step is to coach the
person in how to reach their goal. The commonly
Franke, J., Grünloh, C., Hofs, D., Van Schooten, B., Bondrea, A. and Cabrita, M.
Breaking up Long Sedentary Periods of Office Workers through a Virtual Coach using Activity Data.
DOI: 10.5220/0010721100003063
In Proceedings of the 13th International Joint Conference on Computational Intelligence (IJCCI 2021), pages 389-397
ISBN: 978-989-758-534-0; ISSN: 2184-3236
Copyright © 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
389

available health tracking applications mainly focus on
steps taken per day when it comes to goal setting.
However, as pointed out also by Barbaresko et al.,
a physically active person can generally also have
sedentary behaviour (Barbaresko et al., 2018). In
addition, cramping activity within a small period of
time may not be enough to counteract the damage
done by the preceding sedentary period (Ekelund
et al., 2016). Therefore, to tackle the sedentary
lifestyle during the day at the office, it may be better
to coach a person to get some movement during
prolonged sedentary bouts.
The SMARTWORK project aimed to develop a
worker-centric, artificial intelligence system to fully
support older office workers in sustainable, active
and healthy ageing in their personal and work
environment (Kocsis et al., 2019). One of developed
systems is the healthyMe smartphone application,
that includes various modules (Amaxilatis et al.,
2019). This application is used to monitor
nutrition and weight and it offers office friendly
exercises. Connected to an activity tracker, it
is able to track and visualise physical activity,
sleep and heart rate. In addition, the healthyMe
smartphone application analyses and processes
physical activity for personalised coaching, delivered
through an embodied virtual conversational agent.
The personalised coaching consists of two parts,
based on past data of physical activity: 1) setting
an appropriate, personalised daily activity goal,
and 2) the identification of recurring prolonged
sedentary bouts and daily notification times based
on past data. This paper focuses on the algorithm
behind the second part of the personalised coaching:
the Recurring Sedentary Period Detection (RSPD)
algorithm. This algorithm was designed to detect
recurring sedentary periods per specific weekday,
such that data of non-working days do not interfere
with working days and vice versa, to optimise
the algorithm’s accuracy. The aims of this paper
are 1) to give a detailed description of how the
RSPD algorithm was developed, 2) to evaluate the
accuracy of the algorithm, and 3) to describe how
it was implemented in the healthyMe smartphone
application.
2 DESIGN OF THE RSPD
ALGORITHM
The goal of the RSPD algorithm is to identify
recurring prolonged sedentary periods based on past
physical activity tracked by an activity tracker.
Based on the detected recurring sedentary periods, a
summary is formulated to raise awareness of the user.
Moreover, notification times are pinpointed when the
user should be motivated to get some movement to
break up the sedentary periods.
The RSPD algorithm will become active once two
weeks of activity data are available. Thereafter, it
runs every day to determine daily summaries and
notification times for the upcoming day.
2.1 Identifying Sedentary Periods
A simplified schematic overview of how the sedentary
periods per day are identified is depicted in Figure 1.
This first part of the RSPD algorithm uses a sliding-
Figure 1: A simplified schematic overview of the first part
of the Recurring Sedentary Period Detection algorithm.
window technique to find time intervals in which
the hourly step-goal was not met. The contiguous
time intervals for which the hourly step-goal was not
met are stored and used in the second part of the
algorithm. The first sliding-window starts at the time
the user woke up (t
wake−up
) and slides through the
day in time intervals of 60 minutes, incrementing the
start of the window (t
start
) and end time (t
end
) with 15
minutes until the user went to bed (t
bedtime
). Within
each time interval, it is checked whether steps taken
(sumSteps) meet the hourly step-goal (stepGoal,
250 steps) and if not, the sedentary time interval
(tInAct
start
, tInAct
end
) is saved as sedentary period
under the specific weekday (StoreInAct). Once the
algorithm has iterated throughout the day, it proceeds
to the next available day, until no data is available any
more.
SmartWork 2021 - 2nd International Workshop on Smart, Personalized and Age-Friendly Working Environments
390

2.2 Identifying Recurring Periods
The variable with the stored sedentary periods per
weekday (StoreInAct) is used in the second part of
the RSPD algorithm to identify recurring periods
within the specific weekdays. A simplified schematic
overview of how the identification of recurring
sedentary periods works is depicted in Figure 2.
Figure 2: A simplified schematic overview of the second
part of the Recurring Sedentary Period Detection algorithm.
A time array with a 15 minute interval is initialised
for each day to score when sedentary periods recur
(T Score
day
). The algorithm starts with the first entry
of the stored sedentary periods and increments each
15 minute interval between the start (t
start
) and end
(t
end
) of the sedentary period by one to indicate that
(another) sedentary period has occurred. If present,
the algorithm then proceeds to the next entry of the
stored sedentary periods (StoreInAct) and repeats the
process. After the last available entry, the algorithm
proceeds to the next weekday and repeats the process
until all days of the week are processed.
2.3 Identifying Appropriate Timing for
Coaching
The last step of the RSPD algorithm is to summarise
the sedentary periods per weekday and to pinpoint
the timing when to motivate the user to get some
movement. From the recurring score (T Score
day
),
recurring sedentary periods are identified when at
least 66% of the time this period is determined to
be sedentary. This is determined by dividing the
15 minute interval score (T Score
day
) by the number
of days that are present in the data set. Recurring
sedentary periods of at least one hour are taken into
account in the summary. To prevent lengthy and
complicated summaries, recurring sedentary periods
that are less than one hour apart are considered as one
sedentary period.
Summaries are computed using the start and
end times of the recurring sedentary periods in the
following way: ‘weekdays between t
start
and t
end
,
between t
start
and t
end
and t
start
and t
end
’. Depending
on the number of identified recurring sedentary
periods, this sentence is adjusted such that it can be
directly used in a dialogue with a virtual coach.
To pinpoint when the user should be motivated,
high and low priority recurring sedentary periods are
distinguished. Start time of periods that occur during
100% of the time in the data history are stored under
high priority notification times. The start time of
periods that occur less than 100% and more than 66%
of the time are stored under low priority notification
times. Low priority notification times that fall within
two hours of the high priority notification times are
deleted, to prevent overwhelming the user with too
many notifications. Moreover, it is not desirable to
notify the user close to their usual bedtime and thus,
notifications within two hours of their regular bedtime
are deleted as well.
3 EVALUATION OF THE
OUTCOMES OF THE RSPD
ALGORITHM
To test whether the RSPD algorithm correctly
identifies the recurring sedentary periods, summarises
them and computes correct notification times, a real-
life situation with a typical office worker was tested
(N=1). In the period from 18-02-2021 to 29-04-2021,
a smartwatch (Fitbit Charge HR) was worn 24/7 and
this data was used to evaluate the outcomes of the
RSPD algorithm.
3.1 Methods
To evaluate whether the RSPD algorithm detected the
sedentary periods correctly, cumulative steps taken
over each individual day, taken from the activity
tracker, were visualised. On all daily graphs, all
time intervals computed by the first part of the
RSPD algorithm (Section 2.1) during the day were
shown, either green (hourly step-goal met) or red
(hourly step-goal not met). The resulting identified
sedentary periods were depicted in the graphs as
orange backgrounds. Periods during the day in which
no data was available (non-wear time) were depicted
as grey backgrounds. Visual inspections were done
to evaluate whether the identified orange sedentary
periods from the RSPD algorithm were concurrent at
that moment in time with cumulative step count from
Breaking up Long Sedentary Periods of Office Workers through a Virtual Coach using Activity Data
391

the activity tracker. If this was the case, the algorithm
has identified the sedentary periods correctly.
The recurring sedentary periods were evaluated
using heat maps of consecutive 6-week periods of
data and comparing these to the daily graphs. For
each weekday, a heat map was produced using the
percentages of days where a time interval recurred as
a sedentary period, resulting from the second part of
the RSPD algorithm (Section 2.2). Visual inspections
were done to evaluate whether the percentage of the
recurring periods concurred with the occurrence of
the identified daily sedentary periods (Section 2.1). If
this was the case, the algorithm can correctly identify
recurring sedentary periods. Lastly, these heat maps
were used to evaluate whether the resulting summary
and pinpointed notification times (Section 2.3) were
correctly produced by the algorithm.
3.2 Results
As example of the outcomes of the RSPD algorithm,
the 6-week period between 13-03-2021 and 23-04-
2021 was selected. All Fridays during this period are
shown in Figure 3.
On Friday the 23
rd
of March, between 07:15 and
09:30, the cumulative step count increased from 577
to 668 steps. Within this time period, all of the time
intervals indicated that the hourly step-goal of 250
steps was not met (red). This period of time was
identified by the RSPD algorithm as sedentary. The
data suggested non-wear time after 12:00, in which no
sedentary periods were detected by the algorithm. All
other identified sedentary periods were also inspected
and concurred with the cumulative step count and
time intervals that were indicated as hourly goal not
met. Moreover, in small non-wear periods during
the day (e.g. the Fitbit was charging), see Friday
the 19
th
of March between 13:45 and 16:00, no time
intervals were initiated and no sedentary periods were
identified.
The heat map of identified sedentary periods of the
6-week period is presented in Figure 4. For Friday,
two periods in which all Fridays were indicated as
sedentary were found, between 08:00 and 08:15 and
between 11:15 and 12:00. In Figure 3, these periods
all fall within the identified sedentary periods. This
also holds for all other weekdays and any other
percentage.
The outcomes of the computed summaries and
notification times for the 6-week period are presented
in Table 1. For the Friday, the summary
indicated a recurring sedentary period (more than
66% of the time) between 7:45 and 12:15 and a period
between 19:15 and 23:00. In Figure 4, five separate
recurring sedentary periods can be distinguished,
indicated by orange or red. However, as stated
before in Section 2.3, sedentary periods that are
less than one hour apart should be aggregated into
one sedentary period. Hence, the gaps between
the sedentary periods at 10:00, 10:30 and 19:30
should not be regarded in the summary. Taking this
into account, the summary computed by the RSPD
algorithm was correct. For Mondays, considering
gaps of at least an hour, three separate sedentary
periods can be distinguished. However, the period
between 13:15 and 14:00 was less than an hour and
thus not should not be taken into account in the
summary and notification times. Verifying this in
Table 1, the RSPD algorithm has indeed correctly
computed the summary. All other daily summaries
were inspected as well and were correctly computed.
The notification times did not always concur with
the summaries, see Table 1. For example, on Friday,
the summary started with the first sedentary period
at 07:45, while the first notification time was at
08:00. However, Fridays at 08:00 were indicated
as 100% of the time sedentary and are thus high
priority. Since 07:45 fell within two hours of the high
priority notification, the low priority notification was
suppressed by the RSPD algorithm. On Sunday, there
were only two notification times (09:30 and 14:30),
even though there was a high priority sedentary period
between 21:15 and 22:30. Since the start of this
period was within two hours of the subject’s regular
bed time (23:00), the RSPD algorithm deleted this
notification. All other notifications were verified and
are correctly computed by the RSPD algorithm.
Taken all results into account, the RSPD algorithm
correctly identifies the sedentary periods from
the daily cumulative step count. Moreover, it
accurately identifies the recurring sedentary periods
and computes the summary and notification times
correctly.
4 IMPLEMENTATION OF THE
RSPD ALGORITHM IN THE
SMARTWORK PROJECT
The aim of the SMARTWORK project was to support
older office workers in sustainable, active and healthy
ageing in their personal and work environment
(Kocsis et al., 2019). One of the developed support
systems is the healthyMe smartphone application,
in which the RSPD algorithm was implemented.
Users can connect their activity tracker to the
smartphone application and activity data is used
SmartWork 2021 - 2nd International Workshop on Smart, Personalized and Age-Friendly Working Environments
392

Figure 3: Visualised results of the sedentary period detecting algorithm, for all Fridays during a 6-week period. The boxes
represent the intervals in which the hourly step-goal is either met (green) or not (red). Orange backgrounds indicate sedentary
periods as determined by the algorithm, non-wear time is indicated with grey background.
Figure 4: The results of the Recurring Sedentary Period Detection algorithm based on the 6-weeks period of activity data.
The colour bar represents the percentage of days during that specific interval that are indicated as sedentary.
as described in Section 2. However, instead of
running the RSPD algorithm only once a week,
it was implemented such that it runs in a more
continuous way. Detected recurring sedentary periods
as described in Sections 2.1 and 2.2 are updated every
15 minutes and stored as intermediate result. At the
start of each day, the daily summary and notifications
are generated for that day, by means described in
Section 2.3.
The daily summary is delivered each morning by
the embodied virtual agent, Amelia, in the healthyMe
smartphone application, see Figure 5. Dialogues were
Breaking up Long Sedentary Periods of Office Workers through a Virtual Coach using Activity Data
393

Table 1: Overview of the computed daily summaries and notification times produced by the Recurring Sedentary Period
Detection algorithm. High priority notification times are indicated in bold.
Daily summary Notification times
Monday Mondays, between 07:45 and 12:00, and between 17:45 and 22:30. 08:15 10:45 18:45
Tuesday Tuesdays, between 07:45 and 16:45 08:15 11:30 14:00
Wednesday Wednesdays, between 07:30 and 15:15, and between 18:30 and 23:15. 07:30 20:00
Thursday Thursdays, between 07:45 and 10:45, and between 21:00 and 22:15. 07:45 21:00
Friday Fridays, between 07:45 and 12:15, and between 19:15 and 23:00 08:00 11:15 19:15
Saturday Saturdays, between 16:15 and 22:45 16:30 19:00
Sunday Sundays, between 09:30 and 13:00, and between 14:30 and 22:45 09:30 14:30
Figure 5: Screenshots of the summary produced by the Recurring Sedentary Period Detection algorithm, given through a
dialogue with the virtual coach in the healthyMe smartphone application. Light blue dialogue options indicate the chosen
dialogue option.
written in such a way that the user does not receive
the exact same text each day. The first part of the
sentence, not determined by the algorithm, is chosen
randomly from an available set of options.
During the day, the virtual coach will come back
to notify the user, at the moments determined at
the start of the day, that it is time to get some
movement, for example, through the office friendly
exercises available in the application. Figure 6 shows
a screenshot of the start of the motivating dialogue
with the virtual coach in the healthyMe smartphone
application. Office workers may have fixed meeting
times on some days and thus will be unable to break
up that specific period. However, it would be useful
to get a notification a little before that time, such that
the user can plan to go for a walk before or after the
meeting. Therefore, the notifications are triggered
30 minutes before the start of a recurrent sedentary
period. Moreover, notifications remain available for
the user until the end of the sedentary period. This
way, if an office worker does not immediately open
the notification, the coach can still motivate them
during the length of the sedentary period.
Another important difference between the
RSPD algorithm described in Section 2 and its
implementation in the healthyMe smartphone
application, is that the hourly step-goal is not fixed
(i.e., not 250 steps for each hour each day of the
week). The Fitbit smartphone application uses also
an hourly goal, however, no scientific papers were
found what hourly step-goal is appropriate to reduce
sedentary lifestyles. As some older office workers
SmartWork 2021 - 2nd International Workshop on Smart, Personalized and Age-Friendly Working Environments
394

Figure 6: Screenshot of the start of the dialogue with
the virtual coach in the healthyMe smartphone application
triggered by the pinpointed notification time to break
the sedentary bouts identified by the Recurring Sedentary
Period Detection algorithm.
may lead extreme sedentary lifestyles (<2500 daily
steps) (Tudor-Locke et al., 2011), it may happen that
some days are fully indicated as sedentary period,
leading to only one pinpointed notification time.
Therefore, it was decided to use an outcome of
another algorithm supporting personalised coaching
of physical activity in the healthyMe smartphone
application: the ‘Automated Goal Setting’ algorithm
(AGS). The AGS is used to set daily step goals that
are achievable for the user to prevent demotivation.
It uses past data to flag whether a day of the
week is typically ‘inactive’, ‘normal’, or ‘active’
(compared to a person’s activity level), determines
how many steps the person makes on average for
these classifications and then sets the daily step goals
a bit higher than the average. In the healthyMe
smartphone application, the RSPD algorithm uses the
same flags, but sets the hourly step-goal to 200 steps
(less active days), 250 (normal days) or 300 steps
(active days).
5 DISCUSSION
The aims of this article were to outline the Recurring
Sedentary Period Detection (RSPD) algorithm, to
evaluate its accuracy and to describe how it was
implemented in the smartphone application of the
SMARTWORK project. The evaluation was a N=1
study, with activity data collected over a 10-week
period. Visual exploration of the data confirmed
that the identified sedentary periods correspond to
the time intervals in which the hourly step-goals
were not met (Figure 3). Next, the recurrence of
sedentary periods throughout the week (Figure 4)
were investigated. The recurring sedentary periods
were found to correctly correspond to the identified
sedentary periods in the daily graphs. Lastly, the
daily summaries and notification times (Table 1) were
inspected and determined to be correctly computed
by the RSPD algorithm. Hence, the RSPD algorithm
performs as expected.
During the design of the RSPD algorithm, some
parameters were set based on educated guesses,
namely the initial period before the algorithm
becomes active, the hourly step-goal, minimum
duration of recurring sedentary periods and minimum
interval duration between two sedentary periods. The
two week period before the start of the algorithm
might not be representative of the user’s activity
pattern to provide proper coaching. However, this
period allows for a first impression of the type
of the activity level of the user to initiate the
virtual coaching to encourage the user towards an
active lifestyle. Furthermore, the algorithm will
continuously update and adapt to the routine of the
user throughout time. The hourly step-goal was
chosen based on the step-goal used in the Fitbit
smartphone application. In addition, the choice of
one hour as the shortest sedentary period is in line
with features seen in commercial fitness devices that
send movement reminders every hour (e.g., Fitbit and
Garmin). The decision to aggregate two sedentary
periods with less than one hour in between was taken
to avoid overwhelming the user with notifications.
The RSPD algorithm was designed such that all
above mentioned parameters can be easily adapted by
developers in future works to improve the context of
each specific intervention.
The RSPD algorithm was implemented in one of
the support services of the SMARTWORK project, the
healthyMe smartphone application. This application
uses the outcomes of the RSPD algorithm and delivers
the daily summaries through an embodied virtual
agent in a dialogue. There were differences in how
the RSPD algorithm was described in Section 2 and
Breaking up Long Sedentary Periods of Office Workers through a Virtual Coach using Activity Data
395

how it was implemented in the healthyMe smartphone
application. The notification time was adjusted to 30
minutes prior to the sedentary period. This allows the
user to get some movement before a regular meeting,
or plan in time to do so afterwards. However, if
the user is already doing this regularly, the sedentary
period cannot be broken up, but will still show up
in future daily summaries and notification times.
This may interfere with technology acceptance and
to avoid this, a feature should be implemented in
a future version to allow users to indicate specific
time slots in which movement is not possible (e.g.
recurring meetings). Secondly, the hourly step-
goal was based on the AGS algorithm within the
healthyMe smartphone application that flags days of
the week as typically ‘inactive’, ‘normal’, or ‘active’
days based on past data. These flags were used
to set the hourly goal to 200, 250 or 300 steps,
respectively. While the hourly goals may not be the
same for each day, the goal per flag, however, are still
predetermined and fixed. For example, ‘active’ days
for one person could be 8000 steps, while for another
it is only 3000 steps. In both cases the healthyMe
smartphone application uses 300 steps as the hourly
goal, which would hardly be achievable for the latter
user. Therefore it is likely that for this person most
parts of the day would be indicated as sedentary,
because the hourly step goals have not been reached,
which was supposed to be avoided by having more
personalised hourly step-goals. In future versions of
the RSPD algorithm, it is recommended to determine
averages of the amount of steps taken within the time
intervals that are used. This prevents days from being
identified as one large sedentary period, such that the
most sedentary periods can be targeted.
The RSPD algorithm implemented in the
healthyMe smartphone application will be tested
during the SMARTWORK pilots, with office workers
aged 50+ in the Summer and Fall of 2021. While
the project focuses on older office workers, we
believe the RSPD algorithm will be useful to
support breaking sedentary periods in all ‘couch
potatoes’ with a sedentary lifestyle, from children to
pensioners.
6 CONCLUSIONS
The developed RSPD algorithm correctly identifies
the recurring sedentary periods in physical activity
data. Moreover, it correctly computes summaries
and notification times, that can be directly used in
support services such as the healthyMe smartphone
application. The next step is to evaluate in pilot
studies whether the RSPD algorithm motivates office
workers to break up sedentary periods and to evaluate
the decisions made for the used parameters. These
pilots are planned as part of the SMARTWORK project
in the Summer and Fall of 2021.
ACKNOWLEDGEMENTS
This work is supported by the SMARTWORK project,
funded by the European Commission within H2020-
DTH-2018 (Grant Agreement: 826343).
REFERENCES
Abraham, C. and Michie, S. (2008). A taxonomy of
behavior change techniques used in interventions.
Health psychology, 27(3):379.
Amaxilatis, D., Tsironis, N., op den Akker, H., Cabrita, M.,
and Pardal, A. (2019). Smartwork: D3.2 – first version
of the unobtrusive sensor network.
Barbaresko, J., Rienks, J., and N
¨
othlings, U. (2018).
Lifestyle indices and cardiovascular disease risk:
a meta-analysis. American journal of preventive
medicine, 55(4):555–564.
Ekelund, U., Steene-Johannessen, J., Brown, W. J.,
Fagerland, M. W., Owen, N., Powell, K. E., Bauman,
A., Lee, I.-M., Series, L. P. A., Group, L. S. B. W.,
et al. (2016). Does physical activity attenuate, or
even eliminate, the detrimental association of sitting
time with mortality? a harmonised meta-analysis of
data from more than 1 million men and women. The
Lancet, 388(10051):1302–1310.
Gonz
´
alez-Gross, M. and Mel
´
endez, A. (2013).
Sedentarism, active lifestyle and sport: Impact
on health and obesity prevention. Nutrici
´
on
Hospitalaria, 28(5):89–98.
Kocsis, O., Moustakas, K., Fakotakis, N., Amaxilatis, D.,
Dantas, C., Vassiliou, C., Pardal, A., Quintas, J.,
Hansen, S., Tageo, V., Dougan, P., Vanderheiden,
G. C., Stergiou, A., Hermens, H. J., Cabrita, M.,
and Ziemke, T. (2019). SmartWork: Designing a
smart age-friendly living and working environment
for office workers. In ACM International Conference
Proceeding Series, pages 435–441. Association for
Computing Machinery.
Parry, S. and Straker, L. (2013). The contribution of office
work to sedentary behaviour associated risk. BMC
public health, 13(1):1–10.
Rogers, E. and Wiatrowksi, W. J. (2005). Injuries, illnesses,
and fatalities among older workers. Monthly Lab. Rev.,
128:24.
Rohrer, J. E., Pierce Jr, J. R., and Blackburn, C. (2005).
Lifestyle and mental health. Preventive medicine,
40(4):438–443.
Tudor-Locke, C., Craig, C. L., Aoyagi, Y., Bell, R. C.,
Croteau, K. A., De Bourdeaudhuij, I., Ewald, B.,
SmartWork 2021 - 2nd International Workshop on Smart, Personalized and Age-Friendly Working Environments
396

Gardner, A. W., Hatano, Y., Lutes, L. D., et al.
(2011). How many steps/day are enough? for older
adults and special populations. International journal
of behavioral nutrition and physical activity, 8(1):1–
19.
Breaking up Long Sedentary Periods of Office Workers through a Virtual Coach using Activity Data
397