Activity Recognition for Context-aware Hospital
Applications: Issues and Opportunities for the Deployment
of Pervasive Networks
Jesus Favela & Monica Tentori & Luis A. Castro &
Victor M. Gonzalez & Elisa B. Moran &
Ana I. Martínez-García
Published online: 13 July 2007
# Springer Science + Business Media, LLC 2007
Abstract Hospitals are convenient settings for the deploy-
ment of context-aware applications. The information needs
of hospital workers are highly dependent on contextual
variables, such as location, role and activity. While some of
these parameters can be easily determined, others, such as
activity are much more complex to estimate. This paper
describes an approach to estimate the activity being
performed by hospital workers. The approach is based on
information gathered from a workplace study conducted in a
hospital, in which 196 h of detailed observation of hospital
workers was recorded. Contextual information, such as the
location of hospital workers, artifacts being used, the people
with whom they collaborate and the time of the day, is used
to train a back propagation neural network to estimate
hospital workers activities. The activities estimated include
clinical case assessment, patient care, preparation, informa-
tion management, coordination and classes and certification.
The results indicate that the user activity can be correctly
estimated 75% of the time (on average) which is good
enough for several applications. We discuss how these
results can be used in the design of activity-aware applica-
tions, arguing that recent advances in pervasive and
networking technologies hold great promises for the deploy-
ment of such applications.
Keywords activity estimation . context-aware computing .
hospital activities . neural networks
1 Introduction
Hospitals are good candidates for the introduction of per-
vasive technology [4, 6]. Clinical environments are filled
with increasingly complex technology, including computers
and sensors, where patient care requires coordination and
collaboration among specialists; and the working staff is
highly mobile and technology savvy. Indeed, some ele-
ments of pervasive computing are gradually being intro-
duced in hospitals. These range from wireless networks,
PDAs [9], RFID tags for patient tracking [27], voice-
activated communication devices [32], and sensors for
patient monitoring [29].
One of the challenges of hospital work is the management
of large amounts of information, including patient records,
medical guides, and scientific papers used for evidence-
based medicine [21]. The information needs of hospital
workers are highly dependent on contextual information
such as location, role, time of day, and activity. For instance,
for a nurse attending a patient, the document more relevant
might be the patient’s chart, while for a physician it might be
the medical health record.
Mobile Netw Appl (2007) 12:155–171
DOI 10.1007/s11036-007-0013-5
J. Favela (*) : M. Tentori : L. A. Castro : E. B. Moran :
A. I. Martínez-García
Computer Science Department, CICESE,
Ensenada, Mexico
e-mail: favela@cicese.mx
M. Tentori
e-mail: mtentori@cicese.mx
L. A. Castro
e-mail: luis.castro@acm.org
E. B. Moran
e-mail: elmoran@cicese.mx
A. I. Martínez-García
e-mail: martinea@cicese.mx
V. M. Gonzalez
School of Informatics, University of Manchester,
Manchester, UK
e-mail: vmgonz@manchester.ac.uk

This has motivated the development of context-aware
applications that adapt to changes in the environment to
better assist hospital workers [3, 25]. These applications
focus mostly on supporting intra-hospital communication
and information access based on user location and role. In
this regard, considerable work has been done in the
development of robust approaches to location estimation
for in-door working environments [17]. Although, other
contextual variables such as role and time of day can be
easily determined, estimating the activity being performed
is more complex.
In this work we present an approach for the automatic
estimation of the activity being performed by hospital
workers. This approach is based on the use of a neural
network trained to map from contextual information (e.g.,
location, artifacts being used) to user activity (e.g., clinical
case assessment). The classifier is trained and evaluated with
data captured from close to 200 h of detailed observation and
documentation of hospital workers.
Activity information could be relevant for a diverse
number of hospital applications, such as deciding whom to
call for help or facilitating access to relevant patient
information. For instance, the Vocera communication
system uses a voice-controlled badge to enable mobile users
to communicate over the wireless network currently being
used in some hospitals [32]. The Vocera system enables a
physician to place a call to “a nurse in the emergency unit”.
If the user’s activity could be accurately estimated, a system
such as this one would be able to decide which of the
nurses to call based on their perceived availability.
Similarly, considering that hospital workers, and nurses in
particular, spend a considerable amount of time document-
ing their work, if the system were aware of the nurse’s
activity, for instance that she is administering medication to
the patient at a particular bed, this information could be
automatically captured into the system, simplifying this
laborious task.
Despite all the benefits of activity-aware applications in
hospitals, the development of this type of systems faces
important challenges, such as the design and deployment of
networks of sensors needed to monitor contextual variables
relevant to estimate users’ activities. Recent advances in
wireless communications and electronics have enabled the
development of low-cost, low-power, multifunctional sen-
sor nodes that are small in size and communicate infor-
mation in short distances. These tiny sensor nodes, which
consist of sensing, data processing, and communicating
components, have undoubtedly brought in a new era of
pervasive computing with ubiquitous network connectivity.
Despite the ample gamut of practical, useful application
these pervasive sensors could afford, several issues need to
be addressed for the efficient operation of this technology
in real working settings.
The rest of the paper is organized as follows: In Section 2
we describe a user study conducted in a hospital to deter-
mine the activities performed by hospital staff as well as to
gather the data used to train and evaluate the activity
classifier. Section 3 describes the architecture and training
of the neural network that estimates the activity of users, as
well as the contextual variables used as input for the
classifier. The results obtained are presented and discussed
in Section 4. In Section 5, we present some implications for
the design of activity-aware applications for hospital work.
Section 6 discusses some of the issues that must be solved
for the deployment of networks of pervasive sensors
required to estimate hospital workers’ activities. In section
VI, we describe the previous work related to activity
estimation and how it compares with the approach being
proposed. Finally, Section 7 presents our conclusions and
directions for future work.
2 Activities performed by hospital staff
We conducted a series of workplace studies in a mid-size
public hospital in the city of Ensenada, Mexico. We studied
the Internal Medicine area where more than 70% of the
patients are attended. A preliminary study was conducted in
this area, aimed at revealing the time hospital workers
spend performing different activities; the distance they
move, the places they move to and the reason of doing it;
the people with whom they collaborate more often and the
artifacts they use in support of their work [23]. The main
contribution of this study consists of the characterization of
mobile work and the information management practices
that hospital workers engage in. Then we conducted a more
detailed study to gather additional data. The time of each
action was recorded on paper as it occurred, annotating
details regarding the nature of the actions, artifacts used,
content of conversations, and physical location of individ-
uals. All time stamps were recorded to the second with as
much precision as it was possible. As all records were done
by hand at the field site, they were later transcribed and
integrated into observation reports to facilitate its analysis
and the computation of statistics. Finally, data were grouped
into different classes (i.e., activities of users) which were
later used on the training of the neural network.
Contextual information, such as the location of hospital
workers, artifacts being used, the people with whom they
collaborate and the time of the day, can be used to infer the
user’s activity if one is familiar enough with the hospital’s
rhythms of work. Temporal patterns or “rhythms” are used
by hospital workers to coordinate their activities and
contribute to the regular temporal organization of work in
the hospital [31]. These rhythms are said to be used by
hospital workers to infer the activity conducted by col-
156 Mobile Netw Appl (2007) 12:155–171