Smart home sensors for the elderly: a model for participatory formative evaluation

Abstract

This paper describes an evaluation framework for a smart home project that utilizes sensor technologies to detect activity levels and monitor older adults. The study setting is TigerPlace, an independent retirement community designed to follow the model of aging in place. Technologies used include a stove sensor, a bed sensor, a gait monitor and a video sensor network. The evaluation framework includes focus groups with end users (older adults and health care providers) as well as observations. Preliminary findings indicate an overall positive attitude of older adults towards smart home sensors and a valid and reliable performance. End users need to be included in all stages of a smart home development (design. Implementation and testing) and become actively involved in a formative evaluation of the technology.

Full text

Page 1

Smart Home Sensors for the Elderly: A Model for Participatory
Formative Evaluation
George Demiris, Marjorie Skubic, Marilyn Rantz, James Keller, Myra Aud, Brian Hensel, Zhihai He

Abstract—This paper describes an evaluation framework for
a smart home project that utilizes sensor technologies to detect
activity levels and monitor older adults. The study setting is
TigerPlace, an independent retirement community designed to
follow the model of aging in place. Technologies used include a
stove sensor, a bed sensor, a gait monitor and a video sensor
network. The evaluation framework includes focus groups with
end users (older adults and health care providers) as well as
observations. Preliminary findings indicate an overall positive
attitude of older adults towards smart home sensors and a valid
and reliable performance. End users need to be included in all
stages of a smart home development (design. Implementation
and testing) and become actively involved in a formative
evaluation of the technology.

I. INTRODUCTION
Older adults are living longer and more fulfilled lives,
and they desire to live as independently as possible. But
independent lifestyles come with risks and challenges.
Mobility and cognitive impairments among the elderly lead
to functional decline [1]. Interventions to improve function
include both evidence-based nursing approaches and
innovative technologies. Crucial to successful intervention is
early identification of changing conditions that are
precursors of impairments so that interventions can be
offered at the earliest indications of need. Customized
monitoring can enable the early detection of deteriorating
health conditions, such as a shuffling gait, restless sleep,
rapid change in activity level, or an unusual change in one’s

Manuscript received August 4, 2006. This work was supported in part by
by the National Science Foundation (NSF ITR award IIS-0428420) and the
U.S. Administration on Aging (Grant Nr. 90AM3013).
G. Demiris is with the Department of Biobehavioral Nursing and Health
Systems (School of Nursing) and Department of Medical Education and
Biomedical Informatics (School of Medicine), University of Washington,
Seattle WA, 98195 USA (Phone: 206-381-8613; fax: 206-543-3461; e-mail:
gdemiris@u.washington.edu).
M. Skubic is with the School of Engineering, University of Missouri-
Columbia, MO, USA (e-mail: skubicm@missouri.edu)
M. Rantz is with the School of Nursing, University of Missouri-
Columbia, MO 65211, USA (e-mail: RantzM@missouri.edu)
James Keller is with the School of Engineering, University of Missouri-
Columbia, MO 65211, USA (e-mail: KellerJ@missouri.edu)
Myra Aud is with the School of Nursing, University of Missouri-
Columbia, MO 65211, USA (e-mail: AudM@health.missouri.edu)
Brian Hensel is with the School of Medicine, University of Missouri-
Columbia, MO 65211, USA (e-meail: HenselBK@missouri.edu)
Zhihai He is with the School of Engineering, University of Missouri-
Columbia, MO 65211, USA (e-mail: hezhi@missouri.edu)
typical routine.
To address these issues, researchers are developing "smart
home" technologies to enhance residents' safety and monitor
health conditions using sensors and other devices. Several
pilot projects have introduced smart home technologies for
frail elders in the United States, Europe and Asia:. The
SmartBo project in Sweden, for example, includes a two-
room ground floor demonstration apartment containing
technology for elders with mobility impairments and
cognitive disabilities, such as dementia and developmental
disability. Devices and sensors control lighting, windows,
doors, locks, water outlets, electrical power, stove, as well as
visual and tactile signaling devices, speech synthesizers, and
Braille displays [2](. The PROSAFE project in France
examines devices and sensors that identify abnormal
behavior that can be interpreted as an accident and collect
representative data on patients' nocturnal and daily activity
[3]. The experimental room has been designed to
accommodate patients with Alzheimer’s disease. It is
equipped with a set of infrared motion sensors connected to
either a wireless or wired network. The Hospital Without
Walls is a home telecare project in Australia that includes a
wireless fall monitoring system in which patients wear small
sensors that measure heart rate and body movement. A home
computer records data from several sensors and uploads to a
remote assessment center [4]. Finally, the Aware Home
developed by Georgia Tech in the USA is developed within
two identical independent living spaces allow for controlled
experiments with technology and enable inhabitants to live
on one floor while demonstrating prototypes of assistive
technologies on the other floor [5].
Although these projects have advanced research, most are
primarily demonstrations of technological possibilities or
provide visual displays that are not customized for elderly
users nor do they provide comprehensive information about
activities of daily living but rather graphs and charts that
depict numeric information obtained from the vital sign
assessment. Much work still needs to be done in addressing
fundamental issues on the usability and effectiveness for
older adults. The needs of elders are different from the
general population and the groups typically investigated in
human-computer interaction studies [6,7].
Smart home technology can help keep older adults
independent while controlling costs. However, it is essential
that the solution be driven not by the technological
possibilities but rather by the needs of the older adult
population. Furthermore, it is important to design

Page 2

applications that address the specific needs of their end-
users and follow a model that empowers elders, making
them active participants in the health care process and in the
monitoring of their condition rather than passive recipients
of care services. The aim of this paper is to present an
evaluation framework for a formative and participatory
evaluation of a smart home application. This framework has
been developed and tested with an actual smart home project
that is implemented in a real world setting rather than a
laboratory or demonstration site.

II. METHODS
The study setting is an independent retirement facility in
Columbia, Missouri called TigerPlace. It opened its doors in
2004 and is based on the Aging in Place model as it offers
varied services as needed at the time these are needed rather
than forcing elders to move as their health care needs change
[8]. This model also supports residents to maintain their
health status and activity levels for longer periods of time by
providing ongoing assessment, early illness recognition and
health promotion activities. Ultimately, the aging in place
model aims to help residents avoid expensive and
debilitating hospitalizations and relocation to a nursing
home.
The facility includes 32 independent one and two
bedroom apartment and a wellness clinic staffed by a nurse
three days a week. Furthermore, clinical staff are on call 24
hours a day, 7 days a week. Currently, TigerPlace has 34
residents ranging in age from about 70 to 90 years.
Technology has become an integral part of TigerPlace due
to an ongoing project that investigate the use of sensor
technology to monitor and assess potential problems in
mobility and cognition of elders. The focus is on sensing
alert conditions such as falls, and changes in daily patterns
that may indicate abnormalities. The smart home system that
is in place inclues an in-home monitoring system, an event-
driven anonymized video sensor, and a component that
conducts activity analysis and behavior reasoning.
The In-Home Monitoring System (IMS) [9] consists of a
set of wireless infrared proximity sensors to detect motion,
as well as pressure switch pads (sensor mats) that can be
used to infer specific activities based on the position of the
mat. The system also includes a stove temperature sensor,
switches on cabinet doors and a bed sensor capable of
detecting presence, respiration (normal/ abnormal), pulse
(low, normal or high) and movement in the bed. The Data
Manager collects data from the sensors, date-time stamps the
data and logs it into a file that is sent to a secure server as
binary streams stripped of identifiers. The system exploits
low-cost X10 technologies coupled with specialized filtering
and analysis.
The system includes a passive gait monitor developed by
the University of Virginia that relies on a highly sensitive
displacement sensor [10]. This sensor detects small
deflections in the floor induced by a person walking ten feet
away from the sensor on both carpeted and uncarpeted
wooden and concrete floors. The gait monitor processes the
raw vibration signal, extracts features of significance, and
analyzes the extracted data to provide basic gait
characteristics [10].
The video sensor network complements the IMS by
collecting more detailed information that is not available by
the other sensors. Thus, the video sensor network reduces
false alarms by providing visual information about human
motion. It is called “anonymized” as it uses algorithms to
identify a person in the image and extract a silhouette [11].
The system aims to not only detect emergencies but also
support ongoing monitoring of residents. In this context,
activities are analyzed on multiple time scales. Relatively
short-term observations of events are used to infer activities
such as cooking, getting ready for bed, etc. These short-term
observations are made in the Activity Analysis components.
By observing a sequence of these activities over time, we
can infer a typical daytime pattern of behavior, such as
getting out of bed, cooking, etc. This type of longer-term
behavior reasoning is represented in the Behavior Reasoning
component.
The value of an intelligent monitoring system is to
distinguish a typical pattern for an individual from what
would be an abnormal pattern for this specific individual.
For this purpose we are exploring Hidden Markov Models
(HMMs) for learning and recognizing short-term activity
patterns fusing data from the IMS and the video sensors.
The output of each Activity Analysis process is a descriptor
or a set of descriptors that report the likelihood of an
activity. Fuzzy rules can also be trained when the data is
sufficient, but more importantly, can be modified by the
experts (in this case, the nurses) who can insert specific
domain knowledge. Events that are highly improbable but
nonetheless possible and important to detect are very
difficult to “learn” but easy to incorporate in rules.
The formative evaluation approach described in this
paper involved end-users in both the design and
implementation of the system. End-users are broadly defined
to capture the actual residents, their family members and
health care providers. The design of a usable information
system for healthy users who are familiar with computer
technology is a challenge. When a system needs to address
age-related constraints and the functional limitations of
inexperienced users, it becomes even more difficult. The
aim of the formative evaluation is to increase its functional
accessibility. The term “accessible design” refers to
maximizing the number of potential customers who can
readily use a product [12]. A “functional limitation”
describes a “reduced sensory, cognitive, or motor capability
associated with human aging, temporary injury, or
permanent disability that prevents a person from
communicating, working, playing, or simply functioning in
an environment where other people in the population can

Page 3

function” [12]. For both the design and evaluation of the
displays of activity levels and other monitoring data we
followed design guidelines for elderly patients. Specifically,
we integrated published guidelines for the design of a web-
based clinical monitoring and educational system for elderly
patients [13] into this work. These guidelines cover four
areas: system interface, training, user support, content and
security. The specific guidelines pertain to simplicity, clear
navigation mechanism, selection of appropriate colors,
elimination of distracting features or sound effects, and
inclusion of end users in the design process. The formative
evaluation approach involves the end users in all phases of
the system design and implementation and aims to integrate
feedback assessed via quantitative and qualitative methods
into the design in order to increase usability.
The evaluation model includes observations, focus groups
and interviews with representative users. Focus group
sessions with TigerPlace residents addressed many items,
including the resident's operation of the system and
specifically ease of use, adjusting to changing needs, and
maintenance, impact of the information on lifestyle,
psychological factors, including confidence in using the
interface, reliability of the system, trust in its accuracy and
performance, sense of control and concern for violation of
privacy.
Focus group sessions were also conducted with health
care providers, namely home care nurses of Senior Care, the
agency that operates in TigerPlace, to assess their perception
of visual displays and whether these can be integrated into
the patient monitoring plan, the level of clinical utility and
the extent to which they can be tools for effective
monitoring, early detection and/or prevention of disease and
emergency situations.
Finally, as part of the formative evaluation prlan research
team members made a series of heuristic observations on-
site to identify problems and usability issues of the proposed
applications and the implemented technology. We employed
observations as part of a heuristic evaluation, a usability
inspection method, which refers to a class of techniques in
which evaluators examine an interface for usability issues.
Observation is considered an informal method of evaluating
usability because observers rely on heuristics along with
their experience and the knowledge. The observation
sessions were also useful in validating the sensor
technology. During these sessions residents were asked to
carry out certain activities (for example, meal preparation)
while being observed by two research assistants who made
notes and timed tasks using a portable data assistant (PDA).
The notes were compared with the sensor data to validate
the collected data sets. Furthermore, the observations aimed
to identify whether the installed technology becomes
obtrusive during regular daily activities in the home.

III. RESULTS
We conducted three focus group sessions with a total of
fourteen older adults over the age of 65 (five male and nine
female). Each session lasted approximately one hour.
Overall, participants had a positive attitude towards smart
home technologies. Their perceptions of the potential of the
technology focused on a reactive role (detecting
emergencies) rather than a proactive one (monitoring a
situation to detect trends or predict issues or concerns) [14].
Fall detection was a function that appeared to be uniformly
supported as important. Participants felt that none of the
technologies presented to them would interfere with daily
activities.
Overall, participants saw a balance to be struck between
the benefits of such monitoring, determined by level of
need, and the concomitant perceived intrusion into privacy
at home. Participants also emphasized the need to customize
how the information resulting from the monitoring systems
would be handled.
Furthermore, we conducted one focus group session with
health care providers, namely, four gerontology nurses and
one social worker. The session was facilitated by a member
of the research team. The focus group protocol included
questions about participants’ preferences in accessing
patient related data, their critique of suggested interfaces and
additional questions about types of display and alerts that
will be useful in monitoring and caring for senior residents.
Specific examples of interfaces were displayed and
comments were solicited in terms of advantages and
disadvantages. Participants stated that non-emergency data
sets should be available on a secure website allowing for
providers to access the data at their own discretion and from
several points. Emergency alerts triggered by the system
indicating a situation that requires immediate attention
should be sent in multiple formats, such as email messages,
pager messages, phone calls etc. Participants also provided
specific feedback to examples of interfaces stating a
preference for graphic displays (e.g. pie charts, bar graphs),
use of colors and figures that allow users to zoom in or
obtain more in-depth information if they choose to [15].
All participants expressed preference for a web-based
application that would allow for remote access to the
datasets. Most participants stated that the interface should
allow users to enter interpretations and other notes and
provide a platform for communication with other health care
providers. Visual summaries and overall trends were
perceived as very useful in managing large data sets. Most
participants emphasized the need for internal and external
consistency of the interfaces and interoperability of this
application with other applications and specifically, the
electronic medical record software (regardless of whether
these datasets end up becoming part of the record system).
Finally, we conducted three observation sessions in
apartments of residents who had allowed the installation of
the smart home sensors. Residents were asked to prepare a

Page 4

meal. Two observers took notes and timed tasks during the
sessions without interacting or instructing residents. The
notes were then compared with the datasets that were stored
in the server (triggered from the sensor firings) and in all
cases it could be confirmed that the activities recorded on
paper were also recorded via the sensors (opening and
closing of cupboards, drawers, refrigerators, use of stove
etc.) The aim of the observation sessions was not only to
validate the sensors but also determine whether the different
sensors installed in the kitchen (stove sensor, motion sensors
on refrigerator door, cupboards and drawers) were
interfering with daily activities or were perceived as
obtrusive. In all cases residents confirmed that the sensors
were not noticeable and that they did not change their
routines because of the system.
IV. CONCLUSION
The concept of smart homes places the technology within
the residence in order to increase functionality, security and
quality of life. A smart home enables non-obtrusive
monitoring of residents and there are different levels of
sophistication of the technology involved, ranging from
stand-alone intelligent devices to homes that continuously
monitor residents’ activities and physical status and adopt to
residents’ needs, often providing proactive measures.
Several ethical issues and challenges need to be
addressed, for example, the possibility of such technologies
removing choice and control from the users as they learn to
rely on automation. A further concern is that of transfer of
personal information to third parties, without proper
consent. Specifically, and in relation to the first concern,
there are fears that smart homes may result in a reduction of
social interaction, or may provide tools that substitute
personal forms of care and communication.
In order to address these and other challenges, smart
home applications have to be evaluated on an ongoing basis
and have to include the end-users in all stages. The
evaluation framework presented here was tested in a
residential care facility and demonstrates the potential of
smart homes. Health care providers appreciate the additional
information that sensor systems provide but have specific
expectations of the system interface. Similarly, older adults
have an overall positive perception of smart homes but have
some concerns.
Smart home technologies can potentially revolutionize
home care. Their success depends not only on the technical
feasibility of specific devices and sensors but also on the
level of successful integration into every day life and health
care services. Thus, extensive and ongoing evaluation can
reveal challenges and allow for participants to provide
feedback and improve the overall system.
REFERENCES
[1] AH. Myers, Y. Young , JA. Langlois, Prevention of falls in the
elderly. Bone, (2002),18, 87S-101S.
[2] G. Elger, B. Furugren. 1998, “SmartBo-an ICT and computer-based
demonstration home for disabled people,” Proceedings of the 3rd TIDE
Congress: Technology for Inclusive Design and Equality Improving
the Quality of Life for the European Citizen. Helsinki, Finland June
1998.
[3] M. Chan, H. Bocquet, E. Campo, T. Val, J. Pous, J., 1999, Alarm
communication network to help carers of the elderly for safety
purposes: a survey of a project. Int J Rehab Res, 22: 131-136.
[4] L.S. Wilson, R.W. Gill, I.F. Sharp, S.A. Heitman 2000. Building the
Hospital without Walls- A CSIRO home telecare initiative.
Telemedicine Journal, 6: 275-281.
[5] Kidd, C. D., Orr, R. J., Abowd, G. D., Atkeson, C. G., Essa, I. A.,
Maclntyre, B., Mynatt, E., Starner, T. E., Newstetter, W., 1999, The
Aware Home: A Living Laboratory for Ubiquitous Computing
Research. In Proc. of the Second Intl. Workshop on Cooperative
Buildings – CoBuild ’99. October.
[6] Zajicek, M. 2001. “Interface Design for Older Adults,” In Proceedings
of the 2001 EC/NSF workshop on Universal accessibility of
ubiquitous computing: providing for the elderly, Portugal, May, pp.
60-65.
[7] Gregor, P. and Newell, A. 2001. “Designing for Dynamic Diversity –
Making Accessible Interfaces for Older People,” In Proceedings of the
2001 EC/NSF workshop on Universal accessibility of ubiquitous
computing: providing for the elderly, Portugal, May, pp. 90-92.
[8] MJ. Rantz, Aging in place. Nurseweek, Midwest/Heartland Edition,
(2003), 4(2): 7.
[9] M. Alwan, S. Kell, S. Dalal, B. Turner, D. Mack, R. Felder, In-Home
Monitoring System and Objective ADL Assessment: Validation
Study. Intl. Conf. on Independence, Aging and Disability,
Washington, DC., 2003.
[10] M. Alwan, S. Dalal, S. Kell, R. Felder, Derivation of Basic Human
Gait Characteristics from Floor Vibrations. 2003 Summer
Bioengineering Conference, (2003), June 25-29, Sonesta Beach Resort
in Key Biscayne, Florida.
[11] L. Wang, T. Tan, H. Ning, W. Hu, Silhouette analysis-based gait
recognition for human identification. IEEE Trans. Pattern Analysis
and Machine Intelligence, (2003), 25(12): 1505-1518.
[12] Resource guide for accessible design of consumer electronics—
linking product design to the needs of people with functional
limitations: a joint venture of the electronic industries alliance and the
electronic industries foundation. Telecommunications Industry
Association Web
http://www.tiaonline.org/access/guide.html.
[13] Demiris G, Finkelstein S, Speedie S. 2001. “Considerations for the
design of a web-based clinical monitoring and educational system for
elderly patients,” Journal of the American Medical Informatics
Association. vol.8, no. 5, pp. 468-472
[14] Demiris G, Hensel B, Skubic M, Rantz MJ, Aud MA, Lee J.
Examining senior residents’ willingness to adopt smart home sensors.
Computers Informatics Nursing (Under Review)
[15] G. Demiris, M. Skubic, J. Keller, M. Rantz, D. Parker Oliver, M. Aud,
J. Lee, K. Burks, N. Green, "Nurse Participation in the Design of User
Interfaces for a Smart Home System," in Proc. of the International
Conference on Smart Homes and Health Telematics, Belfast, N.
Ireland, June, 2006, pp. 66-73.
site. Available at: