Cost-effective Design of Real-time Home Healthcare Telemonitoring

Po-Chou Liang and Paul Krause

Department of Computing, University of Surrey, Guildford, U.K.

Keywords: Cost-effectiveness, Healthcare, Telemonitoring, Vital Sign Monitoring, Fall Detection, Movement Pattern

Monitoring, Smart Home, Internet of Things, Mobile Cloud Computing.

Abstract: The importance of telehealthcare for elderly and out-patients has been widely recognized. However, the

adoption rate of home healthcare telemonitoring remains low due to limited evidence for cost-effectiveness.

Our core objective of this work is the cost-effective design of a real-time home healthcare telemonitoring

system based on mobile cloud computing. A second objective is to develop a simulation environment for

evaluating the cost-effectiveness of a telemonitoring system and exploring technology choices. We are at an

early stage, yet the results so far have been encouraging. Whilst we may not be able to deliver a complete

solution, the methodological contribution of test environment plus simulation models will enable us to put

the evaluation of telehealth solutions prior to moving to full-scale trials on a more scientific basis.

1 INTRODUCTION

The rise in both ageing and chronic disease

populations has become a global issue which calls

for a top policy priority to provide proper access to

quality healthcare. Though information and

communications technologies (ICTs) have been used

in almost all aspects of our life, there remains a

considerable question of low adoption rate of remote

healthcare technologies. One of the main reasons, as

indicated by a number of studies (McLean, Prott and

Sheikh, 2011; Limburg et al., 2011), is a lack of

robust evidence for cost-effectiveness.

To address this issue, we set up as our core

objective the cost-effective design of a real-time

home healthcare telemonitoring system based on

mobile cloud computing. Our hypothesis is that the

increasing availability of commodity sensor

technology and computation resource can

dramatically reduce the infrastructure costs of

telemonitoring. In addition, the usability of the

technology is making significant advances -

especially in terms of minimising intrusion on the

patients’ lifestyle (Liang and Krause, 2013).

Our second objective is to develop a simulation

environment in order for us to produce robust

evidence for the cost-effectiveness of a

telemonitoring system so as to explore technology

choices prior to moving to full-scale trials.

Accordingly, a framework based on data from

simulated trials and literature review for conducting

comparative cost-effectiveness analysis is also

proposed. Here, home healthcare telemonitoring is

defined as “the use of ICTs to monitor the vital signs

and activities of in-home patients or elderly

remotely.”

The remainder of this paper is organised as

follows.

In Section 2, we briefly introduce the

development trends in several related areas, such as

telehealthcare, Smart Home and mobile cloud

computing based on literature review. In Section 3,

we present our design and experimental work for the

proposed system. Finally, Section 4 provides our

concluding remarks and future work.

2 LITERATURE REVIEW

AND RELATED WORK

To better understand the development of remote

healthcare, as well as the implications of recent ICT

advances, such as sensor technologies, smart home,

and mobile cloud computing, we have conducted a

broad review of literature in related fields.

2.1 Telehealthcare

The concept of telehealthcare (i.e. the use of ICTs to

provide healthcare remotely) has been explored for

more than thirty years, as evidenced by the

Liang P. and Krause P..

Cost-effective Design of Real-time Home Healthcare Telemonitoring.

DOI: 10.5220/0004723800050015

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2014), pages 5-15

ISBN: 978-989-758-010-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

emergence of nurse call centres in the 1970s in the

UK. As mentioned in Section 1, in recent years, the

problem of ageing and increasing number of people

with chronic diseases have further underpinned the

importance of telehealthcare. Therefore, a great

number of studies on remote home care have

emerged. However, the problem of lacking robust

evidence for cost-effectiveness of related solutions

remains.

A review (Koch, 2006) of the existing scientific

literature on home telehealth during 1990-2003

classified 578 articles from the Medline database as

being relevant to the targeted research field of home

telehealth. Two of the conclusions drawn by this

review were that the impact on those designs for

special user groups, such as elderly, needs to be

further explored, and that in general, evaluation

studies are rare and further research is critical to

determine the impacts, benefits and limitations of

potential solutions.

Another systematic review (Barlow et al., 2007)

identified summaries of 8,666 studies available as of

January 2006 in 17 electronic databases, for

example, the Medline and WTO library. Of those

studies, 98 randomised trials and observational

studies were included in the review. The key

findings included that most studies focused on

people with diabetes (31%) and heart failure (29%),

and that cost-effectiveness of these interventions

was less certain. In addition, there was insufficient

evidence of the effects of home safety and security

alert systems.

Then a systematic review of economic

evaluations (Bergmo, 2010) found only 33 articles

that measured both costs and non-resource

consequences of using telemedicine in direct patient

care. However, the review regarded this as a

considerable increase. It concluded that the

effectiveness measures were more consistent and

well reported than the costings, and that most studies

lacked information about perspective and costing

method.

2.1.1 Cost-effectiveness Analysis

The increasing demand for better healthcare is

manifested in the need to provide better evidence for

informed decision making through economic

evaluation. In this context, Evidence-based Medicine

(EBM), Health Technology Assessment (HTA) and

Comparative Effectiveness Research (CER) have

been used respectively in many organisations to

evaluate the benefits and harms of alternative

treatments, technologies or healthcare deliveries.

Among all techniques of economic evaluations in

healthcare, Cost-effectiveness Analysis (CEA) is

widely adopted.

The National Institute for Health and Clinical

Excellence (NICE) in the UK (2013) defines cost

effectiveness analysis as: “an economic study design

in which consequences of different interventions are

measured using a single outcome, usually in

‘natural’ unit (for example, life-years gained, deaths

avoided, heart attacks avoided or cases detected).

Alternative interventions are then compared in terms

of cost per unit of effectiveness.”

To conduct cost effectiveness analysis, Phillips

(2009) and Muenning (2008) suggested that three

types of costs need to be considered:

 Direct costs: such as drugs, staff time, equipment,

transport of patients;

 Indirect (or Productivity) costs: production losses,

other uses of time; and

 Intangibles: pain, suffering, adverse effects.

The effects of an intervention generally refer to the

changes in patients’ health status. Since there is no

direct way to measure health status, a cost-

effectiveness analysis instead examines patients’

quantity and quality of life with a given health status

(Muenning, 2008). Figure 1 represents the concept

that there are changes in the health status, associated

costs and resulting quality of life and life expectancy

of an observed group of patients having received an

intervention for a period of time

Figure 1: Components of a Cost-effectiveness Analysis,

after (Muenning, 2008).

For independent interventions, the cost-effectiveness

ratio (CER) is calculated to estimate the effects of

different interventions by dividing the costs (C) of

each intervention by its health effects (E) produced,

e.g. life-years gained (LYG) or quality adjusted life

years (QALYs):

CER = C / E (1)

For mutually exclusive interventions, the

incremental cost-effectiveness ratio (ICER) is

calculated by dividing the difference in costs (C)

by the difference in health effects (E) between two

interventions:

ICER =



C / E

(2)

HEALTHINF2014-InternationalConferenceonHealthInformatics

2.1.2 The Whole System Demonstrator

(WSD) Cluster Randomised Trial

In order to better evaluate telecare and telehealth

technologies and their implications for elderly and

people in independent living, the Department of

Health in England (2011) launched the Whole

System Demonstrator (WSD) programme in May

2008. Three sites, Kent, Cornwall and Newham,

were selected to be part of a cluster randomised

controlled trial. With 238 General Practitioners

(GPs) and 6,191 patients with diabetes, chronic

obstructive pulmonary disease (COPD), and

coronary heart disease (CHD), it was believed that

this trial is the world’s largest randomised controlled

trial of telecare and telehealth.

Under this trial, each intervention participant was

given a home unit together with a pendant alarm and

up to 27 peripheral devices for functional monitoring

(such as the home unit and bed and chair occupancy

sensors), security monitoring (such as infrared

movement sensors and property exit sensors) and

standalone devices (not connected to a monitoring

centre, such as big button phones) (Steventon et al.,

2013).

With regard to the key findings of this trial, the

Department of Health in England (2011) announced

that if used correctly telehealth can deliver a 15%

reduction in Accident and Emergency (A&E) visits,

a 20% reduction in emergency admissions, a 14%

reduction in elective admissions, a 14% reduction in

bed days, and a 45% reduction in mortality rates.

However, several in-depth studies on the effect and

cost-effectiveness of this trial reached some

unfavourable conclusions, as in the following:

 Steventon et al., (2012) concluded that, though

both hospital admissions and mortality for

intervention patients were lower, there were no

significant differences between the intervention

group and the control group both in the number

of elective admissions, outpatient attendance,

and emergency visits and in notional hospital

costs to commissioners of care.

 Henderson et al., (2013) found that the QALY

gained by patients using telehealth in addition to

usual care was similar to that by patients

receiving usual care only, and that total costs in

relation to telehealth were also higher. As such,

this study concluded that telehealth does not

seem to be a cost effective addition to standard

support and treatment.

 Steventon et al., (2013) concluded that telecare

did not significantly reduce the use of health and

social care services.

Another study (Sanders et al., 2012) identified that

concerns about both competency to operate

equipment and threats to identity, independence and

self-care (which might be undermined, among

others, by not getting outside, but doing monitoring

indoors even on holidays) are two of the main

barriers to adoption of telehealth and telecare

interventions within this trial.

Based on the abovementioned findings, we

consider that the WSD trial could serve as an

important reference for conducting cost comparison,

selecting inexpensive technologies, devising proper

service models and designing workable system

architecture for the proposed home healthcare

telemonitoring system.

2.2 Smart Home and Internet

of Things

The concept of the so-called “smart home” or

“digital home” has been proposed for more than a

decade, aiming to transform our home environment

into an intelligence-embedded living space. This

paper uses these two terms alternately. According to

Elderly Accommodation Counsel (2003), the UK’s

Department of Trade and Industry’s “Smart Homes

Project” defined smart home as “A dwelling

incorporating a communications network that

connects the key electrical appliances and services,

and allows them to be remotely controlled,

monitored or accessed.”

There were several industrial initiatives for smart

home driven mainly by manufacturers and network

providers. For instance, the Open Service Gateway

Initiative (OSGi) Alliance founded in 1999 focuses

on open specifications for remote management and

the delivery of services into the home. At almost the

same time, the Konnex Association was formed in

1999 to promote an open standard, called KNX, for

home and building control. Other similar effort

included the establishment of Universal Plug & Play

(UPnP) Forum in 1999 and the Digital Living

Network Alliance (DLNA, originally named

“Digital Home Working Group”) in 2003.

Since their inception, these industrial initiatives

have gradually expanded and gained wider support

across different industrial sectors and players. For

example, today PS2, XBOX 360 and personal

computers with MS Windows 7 installed all support

DLNA standards, and both OSGi and DLNA

specifications support UPnP standards. However, on

the service side, the market has only developed to a

very limited extent.

With regard to smart home monitoring systems,

Cost-effectiveDesignofReal-timeHomeHealthcareTelemonitoring

Gaddam, Mukhopadhyay and Sen Gupta (2011)

stated that when more sensors are added to a smart

home system, the system becomes complicated to

handle and the maintenance becomes a challenge. In

our opinion, this is also quite true to home

healthcare telemonitoring, as remote home

healthcare is generally considered as a subcategory

of smart home (Wu et al., 2009).

Figure 2: The Concept Diagram of a Typical Smart Home

System.

Generally, a home gateway or a control hub (see

Figure 2) interconnects one or more home networks

and the Internet/access network (sometimes a cable

modem or router is also needed), and controls other

in-home devices and sensors (den Hartog, et al. 2004;

Wei et al., 2010). For a commercialised smart home

service package, the central server, i.e. one or a

group of computers, is usually located at the service

provider’s premise. However, in other cases it is

common to see that the proposed system

architectures require one or multiple servers (or

called controllers) to be set up within the smart

home environment, one for each platform that is

being used by a controlled device (Zimmermann1

and Vanderheiden, 2007). From our viewpoint, this

kind of design would increase the complexity of

system installation and maintenance.

One important evolution of recent ICTs, which

has great implications for the development of smart

home, as well as home healthcare telemonitoring, is

the emergence of the Internet of Things (or IOTs).

The International Telecommunication Union (ITU)

(2005) described the IOTs as a new form of

communication between people and things, and

between things themselves, which “connects the

world’s objects in both a sensory and an intelligent

manner.”

The basic architecture of the IOTs consists of

three layers: application layer, network layer and

sensor layer (Kang et al., 2011), which in our

opinion can be naturally fitted into the concept

framework of a smart home system as depicted

previously in Figure 2.

According to the Cluster of European Research

Projects on the Internet of Things (CERP-IoT), a

large number of application domains in the field of

IoTs have been identified (Sundmaeker et al., 2010).

We believe that among others, Intelligent Buildings,

Healthcare (monitoring of parameters, positioning,

real time location systems), Independent Living

(wellness, mobility), and Environment Monitoring,

are all applicable to supporting our envisioned smart

home, as well as home healthcare telemonitoring.

A 2011 study (McCullagh and Augusto, 2011)

investigating the potential of IoTs to monitor health

and wellness concluded that the underlying

technology is available but needs to be turned into a

solution which can become pervasive in society.

This is the gap that this research intends to fill by

using low-cost, off-the-shelf technologies to build up

evidence for a solid solution.

2.3 Sensor Technologies

As mentioned in Section 2.1, sensors form an

indispensable component of a smart home system, as

well as a healthcare telemonitoring system. In

general, a sensor is capable of detecting three but

intrinsically related categories of events (Faludi,

2010):

 Direct or proximal phenomena: events that

directly trigger the sensor device;

 Indirect or distal phenomena: remote causes of

the local events actually triggering the sensor;

and

 Context and subtext: the situation surrounding an

event.

However, contextual information inferred from both

direct and indirect phenomena might still involve

some degree of uncertainty. This demonstrates the

importance of a well-designed event reasoning

algorithm that can increase the accuracy of context

inference based on a limited set of monitored data.

Today there are a great variety of electronic

sensors available in the marketplace. In the field of

telehealthcare, there are also increasing focuses on

the development of the so-called Body Sensor

Networks (BSNs) for on-body applications. For the

purpose of our cost-effective design of home

healthcare telemonitoring, this research pays special

attention to existing inexpensive, portable and easy-

to-use sensor technologies/platforms.

2.3.1 ZigBee

ZigBee is a standards-based low-power wireless

Home Gateway

/Control Hub

Device 1

Internet

Central

Server

Remote Control

Panel/Context

Status Checking

Device 2

Device n

HEALTHINF2014-InternationalConferenceonHealthInformatics

technology mainly operating in the 2.4GHz radio

frequency band. It is based on the IEEE 802.15.4

standard with add-on network and security layers

and an application framework. The ZigBee Alliance

was established in 2002 to develop relevant

specifications and to promote ZigBee standards

adoption. Today, the ZigBee Alliance has over six-

hundred certified products, ranging from home

appliances, energy efficiency apparatuses,

networking devices, to health and fitness sensors.

ZigBee Health Care was introduced to provide an

industry-wide standard for exchanging data between

a variety of medical and non-medical devices

(ZigBee Alliance, 2009).

Based on different topologies, such as pair, star

and mesh, a ZigBee sensor network consists of one

coordinator node and at least one router or one end-

device node (Faludi, 2010). In a ZigBee network,

each node can communicate with all the others by

way of its nearest neighbour so that only small

amount of power is needed for radio transmission.

In addition, with the embedded capability to perform

self-healing, a ZigBee mesh network can reconfigure

itself and route around a problem area when some

nodes are failed or removed. Other important

features of Zigbee 2012 specifications (ZigBee

Alliance, 2012) include data security based on

Advanced Encryption Standard (AES), low-power

consumption for better battery life, and low cost in

comparison with other wireless technologies.

At the time of this writing, there are only a few

kinds of sensors available in the health and fitness

sub-category. Besides, ZigBee’s limited

programming capacity to perform software

logic/data processing suggests that all raw data

needs to be dealt with by other layers in a smart

home or IoT system. This would result in a greater

amount of data traffic and lower data reliability.

2.3.2 Arduino

Arduino is an open-source microcontroller platform

for physical computing. It was originally designed in

2005 to provide students with an inexpensive

microcontroller to drive their robotic projects. To

date, it has evolved into a popular tool kit for

prototyping and do-it-yourself work.

By attaching different combinations of various

sensors and actuators to a programmable

microcontroller board, many different tasks, such as

environmental (e.g. temperature and humidity)

monitoring and home automation (e.g. door/window

opening), can be performed in a way that is based on

the user-uploaded software programme. There are

also a number of different communications modules,

such as serial port (e.g. USB), Wi-Fi, Bluetooth, and

web server, available for use to transmit the

programmed outputs, such as the status of the board

and/or the monitored data, to other devices or a web

client.

According to Arduino website (2013), the main

advantages offered by Arduino include: low-cost as

compared with other microcontroller platforms,

cross-platform (among MS Windows, OS X, and

Linux), simple programming environment, and open

source with extensible software and hardware. From

our point of view, the capabilities both to conduct

on-board data processing by the microcontroller to

provide more reliable and meaningful monitored

data, and to interconnect and interoperate with a

variety of devices, such as smartphone and ZigBee,

are two other important features that enable Arduino

to provide more flexible sensory solutions.

2.4 Mobile Cloud Computing

and Home Healthcare

Along with the recent prevalence of smart mobile

devices in our daily life, hundreds of thousands of

available mobile applications, or the so-called

“Apps” are targeting a great diversity of consumer

segments. According to Sarasohn-Kahn (2010), “as

of February 2010, there were 5,805 health, medical,

and fitness applications in the Apple AppStore. Of

these, 73% were intended for use by consumer or

patient end-users, while 27% were targeted to health

care professionals.” There were also Apps using

available sensors, including accelerometers, infrared

photo-detectors and glucometers, for home

measuring. These figures and developments

represent both challenges and opportunities to this

research.

Meanwhile, both mobile computing and mobile

cloud computing have recently gained increasing

attention from ICT researchers and developers.

According to Huang (2011), mobile computing

research refers to the study on how mobile devices

learn their own status and surrounding contexts to

better support mobile applications.

Regarding mobile cloud computing, there are

two different viewpoints (Qureshi et al., 2011). One

refers mobile cloud computing as making use of

cloud resources, such as computing power and

storage, to help perform tasks or store data from

mobile devices, which generally have limited

computing capacity and data storage. The other

recommends that with mobile cloud computing, both

data processing and storing be done by the mobile

Cost-effectiveDesignofReal-timeHomeHealthcareTelemonitoring

device. For the purpose of this research, we take

both views to give a broader definition of mobile

cloud computing. With this, it is apparent that by

adopting mobile cloud computing, an application

can be further empowered by mobility together with

the main advantages, such as on-demand self-

service, resource pooling, rapid elasticity, and pay-

per-use, derived from cloud computing.

Some sophisticated system architectures have

been proposed so that several mobile devices can

work together to perform a particular task or each

device can provide its remaining resources for other

devices. For example, Hyrax (Marinelli, 2009), a

mobile cloud computing platform, was proposed

based on Hadoop to provide data sharing and

distributed data processing among a group of

networked mobile phones. In its implementation,

Hyrax used two conventional computers to perform

Hadoop-related NameNode and JobTracker

processes and cluster initialisation. Clearly, such

complex system architecture would not fit into our

requirements.

Cheng and Zhuang (2010) proposed a Bluetooth-

enabled, in-home patient monitoring system for

early detection of Alzheimer’s disease. The

proposed system required every room in the house to

be equipped with a Bluetooth access point (AP), and

all APs needed to be connected to a local database

(i.e. a laptop). A Bluetooth-enabled pocket PC was

carried by the target person in the house and tried to

find an AP with strongest signal to which to connect.

If the target entered another room, the pocket PC

would try to connect to another AP. By such an

approach, the movement pattern of the target could

be identified and stored in the local database. The

data could then be transmitted to a remote medical

practitioner for diagnosis, or be analysed by an

assumed decision engine to see if the target had any

early signs of Alzheimer’s disease. From our

perspective, this proposal was not very practical, as

both the physical locations of each AP and the

layout of the house would seriously affect the

detectable Bluetooth signal strength and in some

cases would even cause failures in establishing

Bluetooth connection. Accordingly, the deployment

of Bluetooth APs could be very complex.

MoCAsH (Hoang and Chen, 2010) was a

proposed mobile cloud for assistive Healthcare. Its

system architecture included (i) sensors and mobile

agents, (ii) a context-aware middleware, (iii) a

collaborative cloud, and (ix) a cloud portal. The

cloud portal allowed authorized users to access

offered services, including checking sensor status,

updating context-aware rules, and accessing back-

end cloud platform management centre. It also

proposed a P2P federated cloud model to schedule

distributed clouds and their resources, and to

enhance data security. In our view, this project could

have served as a good reference for our prototyping

and design. However, it put its main focus on how to

integrate mobile devices into a federated cloud

architecture without addressing how to implement

non-built-in sensors’ deployment and patient

monitoring.

Wang et al. (2008), as well as Yang and Zhao

(2011), proposed to place a tri-axial accelerometer at

the head level with a pre-defined position and angle

to detect human falls. In our opinion, this kind of

physical setting is not only impractical in home

patient/elderly monitoring, but also intrusive to the

monitored people. Viet, Lee and Choi (2012) used

an Android smart phone which has a built-in

accelerometer and an orientation sensor to perform

human fall detection. It was concluded that the

proposed system reached 85% accuracy in 260 trials.

However, since the implementation was based on a

standalone mobile phone, the proposed system did

not possess any remote monitoring capability.

3 RESEARCH DESIGN

AND CURRENT WORK

3.1 Requirements and Considerations

Based on the preceding technology reviews we have

identified the following set of requirements for a

cost effective telemonitoring system.

3.1.1 Functional Requirements

 Vital sign monitoring: This refers to the on-

demand monitoring of patients’ vital sign

parameters, such as body temperature, heartbeat

rate, oxygen in blood, blood pressure, blood

glucose, cardiogram, and sweat level.

 Safety monitoring: The main function will be

real-time human fall detection with alerts being

sent automatically to designated caregivers

and/or healthcare professionals via a healthcare

dashboard, SMS and video phone.

 Emergency call-for-help tool kits: This refers to

the provision of a portable alarm; once pressed

by a patient, it would send out an alert to

designated caregivers and healthcare

professionals via the healthcare dashboard with

configurable, automated SMS and video phone

call out and call in functions.

HEALTHINF2014-InternationalConferenceonHealthInformatics

 Activity monitoring: This includes movement

pattern monitoring, bed and chair occupancy

sensing and property exit sensing for social care

purposes.

 Service portal/management console and

healthcare dashboard: The service

portal/management console allows authenticated

in-home patients, as well as remote caregivers

and healthcare professionals to control and

manage the sensors. It also allows them to set

their preferences and care plans for healthcare

monitoring, as well as to manage and access

context/health data via a healthcare dashboard.

 Authenticated database management and access:

This refers to a database system that provides

authenticated users with remote management and

access to the large volume of monitored data.

3.1.2 Basic Considerations

The following considerations with criteria for

evaluation need to be addressed throughout the

whole system development life cycle to ensure that

the research objective can be successfully fulfilled.

 Low-cost: There should be no significant amount

of capital expenditures (Capex) and operational

expenditures (Opex) on system setup and

operations.

 Easy-to-deploy-and-use: In general, the end-

users, especially those living independently,

should be able to set up and operate the system.

 Less intrusive: Generally, the monitoring should

not hinder patients’ normal daily routine and

mobility.

 Robust enough: The system should embrace

fault-tolerant and resilient design to maximise

service availability. When the Internet is not

available or the cloud side is unreachable, the

application on the mobile device as well as the

monitoring task should be able to continuously

function properly.

 Security and data privacy: The system should

employ proper access control, user

authentication, data encryption, and secured data

transmission to enhance data privacy and

security.

 No vendor lock-in: The system design should

avoid or at least minimise the impact of vendor

lock-in issue by taking the portability of each

monitored patient’s data into account.

 Good performance and elasticity: The system

performance and elasticity need to be well

managed to provide streamlined user experience

and service provision.

3.2 System Design

Figure 3: High-level System Architecture.

As shown in Figure 3, the proposed system

architecture for the home healthcare telemonitoring

system consists of four main modules, i.e. Sensor

Nodes, User Agent(s), Service Gateway (Cloud

Broker) and Public Cloud(s). The main functionality

of each module is illustrated below:

 The User Agent(s) Module: Its main functions

include: (i) a user interface for users to control

and manage the sensors, to set their preferences

and care plans for healthcare monitoring and to

manage context and health data; (ii) an

intelligent data aggregator that connects with a

variety of sensors, collects real-time sensor data

through high-level sensor APIs and transmits it

to cloud storage, and performs context/health

data reasoning based on preset parameters and

algorithms to automatically trigger an alert; and

(iii) a portable personal healthcare assistant that

can work with, and without an Internet

connection (Liang and Krause, 2013). Figure 4

illustrates the architecture diagram of this

Module.

 The Service Gateway Module: Its essential

functions include: (i) a management console both

for performing administrative tasks and for

providing caregivers and healthcare professionals

with a service portal for remote patient data

access and alert notification via a healthcare

dashboard; and (ii) a cloud manager/broker that

performs protocol translations for requests and

responses between the User Agent(s) and the

Public Cloud(s), and allocates cloud resources

based on user preferences or performance

criteria.

Cost-effectiveDesignofReal-timeHomeHealthcareTelemonitoring

Figure 4: Architecture Diagram of the User Agent Module

(Liang and Krause, 2013).

 The Public Cloud(s) Module: With the help of

the Service Gateway Module, this Module can

consist of a variety of cloud platforms, such as

AWS, GAE, cloud-based Social networking

websites, and free cloud-based health data

storage, e.g. HealthVault.

Figure 5: Architecture Diagram of the Sensor Nodes

Module.

 Sensor Nodes: This module is composed of a

number of off-the-shelf portable sensor devices

(see Figure 5) to collect data for vital sign

monitoring, safety monitoring and activity

monitoring.

3.3 Experimental Design

and Cost-effectiveness Analysis

3.3.1 Limitations

Due to limited resources, time and funding in

particular, it is impractical for this research to design

and implement a randomised controlled trial to

measure costs and effects over several years, as

normally done in the health sector. Instead, this

research will only conduct some simulated trials and

adopt a revised comparative effectiveness analysis

approach for economic evaluations. The purpose

behind this is to evaluate whether there is a case for

designing a full scale trial without committing to the

expense of such a trial.

Another limitation is the unavailability of low-

cost, portable, programmable, and, most

importantly, clinically certified, sensor devices in

healthcare. As a result, this research will have to use

uncertified sensor devices, making a real clinical

trial unrealisable.

3.3.2 Experimental Design and Results

To date, a proof-of-concept prototype using Ruby on

Rails framework has been developed mainly based

on expanding and integrating three standalone

projects under the same theme of “Medical Alert

Management” in the Department of Computing,

University of Surrey, UK. Each of them had

different focuses, ranging from data presentation,

sensor data collection, to data storage.

Meanwhile, the development of the Sensor

Nodes Module and the integration of a real-time

remote monitoring function and those three projects,

as well as the implementation of the User Agent

Module on iPhone 5, are underway. Currently, by

using a web browser, an authorised remote user such

as a registered GP can use the dashboard to access

and review historical patient monitored data stored

in a remote server’s MySQL database. In addition,

the user can switch some panels inside the

dashboard to display dynamic real-time monitored

data, such as body temperature, heartbeat rate and

ambient temperature, which is first received by the

User Agent Module through either a Bluetooth

wireless connection or wired.

For the purpose of human fall detection, we

currently adopt a wearable device approach, mainly

based on accelerometry-related parameters, such as

the sum vector (SV) of acceleration in X-Y-Z axes

(see Equation 3). Figure 6 shows both SV and

acceleration signatures in an intentional forward fall

using the on-board accelerometer of the Texas

Instruments’ SensorTag.

222

zyx=SV 

(3)

When building our fall detection algorithm, we first

assumed that a fall followed by lying motionless is

an emergency that needs to trigger an alert. 30

HEALTHINF2014-InternationalConferenceonHealthInformatics

simulated activities of daily living (ADL), each

followed by an intentional forward fall on a cushion,

were performed by locating either a SensorTag or an

iPhone with a built-in accelerometer at different

places of a volunteer’s body, such as ear side, jacket

pocket, shirt chest pocket, pants pocket, or handheld

(Liang and Krause, 2013). To make our simulated

falls closer to reality, we did not strictly confine the

sensors to a certain tilting angle or orientation. Such

a research design is apparently different from a

number of studies (Kangas et al., 2007; Yang and

Zhao, 2011; He, Li and Bao, 2012).

Figure 6: Changes of SV and Acceleration in X-Y-Z Axes

in an Intentional Forward Fall.

The results from 22 falls (eight falls were

excluded due to noisy data) revealed that when SV

first drops below 0.79g (1

threshold) before

bouncing over 1.48g (2

threshold) and then after a

few oscillations it remains in the interval between

1.125g and 0.89g (3

threshold) for more than 2

seconds, a serious fall might have occurred.

Nevertheless, dropping or throwing an

accelerometer could produce similar SV signature.

Consequently, we add another threshold at 0.15g (4

threshold) to detect a free fall situation, which

enables us to distinguish all device drops/throws

from human falls.

Table 1: Results of fall detection using 3-threshold or 4-

threshold algorithms (Liang and Krause, 2013)

(accelerometer range: ±2g, sampling rate: 10Hz).

3 thresholds 4 thresholds

Sensitivity 95.5% 95.5%

Specificity for device

drops/throws

0% 100%

Specificity for ADLs 95.5% 95.5%

In Table 1, sensitivity is defined as the

percentage of successfully identified falls and

specificity is the percentage of successfully

identified non-fall tests. Indeed, we have also

developed another algorithm to identify intentional

device shaking events, which sometimes can

produce almost identical SV signatures to human

falls. However, instead of using the new algorithm at

the expense of less sensitivity, we add a function to

ask for user confirmation before an alert is sent to

remote caregivers.

Regarding vital sign monitoring, we use an

Arduino-compatible platform (Seeeduino Stalker

v2.1 shield manufactured by Seeed Studio) and

clinically uncertified sensors (e-Health Sensor

Platform v1.0 with optional sensor kits, such as

pulse, oxygen in blood, body temperature and body

position sensors by Cooking Hacks). Nevertheless,

the accuracy and reliability of the used sensors have

been disappointing so far. For example, the highest

body temperature measured by the e-Health Sensor

Platform’s thermometer was under 30 degree

Celsius and the body position sensor just did not

work. According to the manufacturer of the e-Health

Sensor Platform, a possible reason might be

incompatibility between the e-Health Platform and

the Seeeduino Stalker shield, as the former is

designed for Arduino. However, after some

relatively minor modifications to the sensors and

wiring, our User Agent Module can start receiving

meaningful data from some of the sensors. We

believe the results can be further improved with

more work (Liang and Krause, 2013).

As for movement pattern monitoring, we plan to

use received signal strength from three triangular

deployed reference sensors, such as SensorTags, for

in-home patient location and movement estimation.

However, due to limited resources, we currently

have only one SensorTag. By measuring received

signal strength from a man-carried SensorTag, we

can roughly estimate the distance between the man

and the User Agent with an accuracy of around two

to three meters.

3.3.3 Cost-effectiveness Analysis

Due to the above-mentioned limitations, this

research calculates neither CER nor ICER directly,

but performs simulated trials to predict the

effectiveness of the proposed system and then

conduct cost-effectiveness analysis based on a

revised comparative effectiveness analysis approach.

Four types of comparisons, including the

comparison between simulated control and

intervention groups, for building up evidence of

cost-effectiveness are made.

Figure 7 shows the concept diagram for a

comparative effectiveness analysis approach which

compares our simulated trials with existing

randomised controlled trials. Data about the costs

and effects (the resulting changes in a group of

patients’ health status from Health Status X to

Cost-effectiveDesignofReal-timeHomeHealthcareTelemonitoring

Health Status Y) of a known Intervention Y is first

obtained from literature review. Then we can claim

that our proposed Intervention Z can provide the

same QALY effects or better QALY effects (i.e.

Health Status Y+ with Quality of Life Y+ and Life

Expectancy Y+) than Intervention Y, if Intervention

Z has the same or better functionality/performance.

Finally, Cost Z and ICER of Intervention Z are

calculated for cost-effectiveness analysis (Liang and

Krause, 2013).

Figure 7: Concept Diagram for a Comparative

Effectiveness Analysis Approach (Liang and Krause,

2013).

To enable ourselves to satisfactorily conduct

cost-effectiveness analysis and to make claims about

the generalisability of this research, we first need to

further improve the reliability and accuracy of both

our event reasoning algorithms and the sensors. The

technical problem of incompatibility among devices

also needs to be better resolved. Meanwhile, a more

stable and well-defined testing environment has to

be carefully designed to make our simulation more

meaningful and robust.

4 CONCLUSIONS

In this paper, we have discussed the long-standing

problem of lacking robust evidence for cost-

effectiveness of healthcare technologies. To tackle

this issue, we have broadly assessed the implications

of recent advances in sensor technologies, smart

home, Internet of Things and mobile cloud

computing in support of achieving a cost-effective

design of a home healthcare telemonitoring solution.

We then have proposed a system architecture based

on mobile cloud computing and developed a proof-

of-concept prototype together with a novel

comparative cost-effectiveness analysis approach

based on simulated trials. Through the experimental

design, we believe that the proposed system is a

good foundation for moving forward.

In addition to the future work mentioned in

Section 3.3.3, we will also work on the development

of the Service Gateway Module to integrate all the

proposed system components as a whole, and

complete simulated trials and cost-effectiveness

analysis. Whilst we may not be able to deliver a

complete solution, we are confident that the

methodological contribution of test environment

plus simulation models will enable us to put the

evaluation of telehealth solutions prior to moving to

full-scale trials on a more scientific basis.

REFERENCES

Arduino website, 2013, “Arduino – Introduction”, viewed

28 Apr. 2013, <http://arduino.cc/en/Guide/

Introduction>.

Barlow, J., Singh, D., Bayer, S. & Curry, R., 2007, “A

systematic review of the benefits of home telecare for

frail elderly people and those with long-term

conditions”, Journal of Telemedicine and Telecare,

vol. 13, no. 4, pp. 172-179.

Bergmo, T., 2010, “Economic evaluation in telemedicine

– still room for improvement”, Journal of

Telemedicine and Telecare, Vol. 16, no. 5, pp. 229-

331.

Cheng, H.T. & Zhuang, W., 2010, “Bluetooth-enabled in-

home patient monitoring system: Early detection of

Alzheimer's disease”, IEEE Wireless

Communications, Vol. 17, Issue 1, pp.74 – 79.

den Hartog, F. T. H. et al. 2004, “Convergence of

Residential Gateway Technology”, IEEE

Communications Magazine, May 2004, pp. 138-143.

Department of Health, 2011, “Whole System

Demonstrator Programme: Headline Findings”.

Elderly Accommodation Counsel, 2003, “Smart Home - A

Definition”, viewed 5 Jan. 2013,

<http://www.housingcare.org/information/detail-2545-

smart%20home-a-definition.aspx>.

Faludi, R., 2010, “Building Wireless Sensor Networks”,

O’REILLY, CA, USA.

Gaddam, A., Mukhopadhyay, S.C., Sen Gupta, G., 2011,

“Trial & experimentation of a smart home monitoring

system for elderly”, IEEE Instrumentation and

Measurement Technology Conference (I2MTC).

He, Y., Li Y. & Bao, S.D., 2012, “Fall Detection by Built-

In Tri-Accelerometer of Smartphone“, IEEE-EMBS

International Conference on Biomedical and Health

Informatics (BHI 2012), Conference Proceedings, p.

184-187.

Henderson, C. et al., 2013, “Cost effectiveness of

telehealth for patients with long term conditions

(Whole Systems Demonstrator telehealth

questionnaire study): nested economic evaluation in a

pragmatic, cluster randomised controlled trial”,

BMJ2013;346:f1035 doi: 10.1136/bmj.f1035.

Hoang, D. B. & Chen, L., 2010, “Mobile Cloud for

HEALTHINF2014-InternationalConferenceonHealthInformatics

Assistive Healthcare (MoCAsH)“, 2010 IEEE Asia-

Pacific Services Computing Conference (APSCC),

Conference Proceedings, pp.325 – 332.

Huang, D., 2011, “Mobile Cloud Computing”, Multimedia

Communications Technical Committee, IEEE

Communications Society, E-Letter, Vol. 6, No. 10, pp.

1-4.

International Telecommunication Union (ITU), 2005,

“ITU Internet Reports 2005: The Internet of Things”.

Kang, B. et al. 2011, “Design of an Internet of Things-

based smart home system”, the 2nd International

Conference on Intelligent Control and Information

Processing, Conference Proceedings, pp.921-924.

Kangas, M. et al., 2007, “Determination of simple

thresholds for accelerometers for fall detection”, The

29th Annual International Conference of the IEEE

EMBS”, Conference Proceedings, p.1367-1370.

Koch, S., 2006, “Home telehealth – Current state and

future trends”, International Journal of Medical

Informatics, Volume 75, Issue 8, Pages 565–576.

Liang, P.C. & Krause P., 2013, “Cost-effective Healthcare

Telemonitoring”, Conference on Mobile and

Information Technologies in Medicine 2013

(MobileMed), Conference Paper 4.

Limburg, M., Gemert-Pijnen, J., Nijland, N., Ossebaard,

H., Hendrix, R. & Seydel, E., 2011, “Why Business

Modeling is Crucial in the Development of eHealth

Technologies”, Journal of Medical Internet Research

(JMIR), 13(4): e124.

Marinelli, E., 2009, “Hyrax: Cloud Computing on Mobile

Device Using MapReduces”, Carnegie Mellon

University, MSc Thesis.

McCullagh, P. & Augusto, J., 2011, “The Internet of

Things: The Potential to Facilitate Health and

Wellness”, CEPIS UPGRADE, Vol. XII, Issue no. 1,

pp. 59-68.

McLean, S., Prott, D. & Sheikh, A., 2011, “Telehealthcare

for long term conditions”, BMJ 2011; 342:d120, doi:

10.1136/bmj.d120.

Muenning, P., 2008, “Cost-Effectiveness Analysis in

Health: A Practical Approach”, 2

ed., Jossey-Bass,

Hoboken, NJ, USA.

National Institute for Health and Care Excellence, 2013,

“Guide to the methods of technology appraisal 2013”.

Phillips, C., 2009, “What is cost-effectiveness?”, Health

economics, 2

ed., viewed 17 Apr. 2013,

<http://www.medicine.ox.ac.uk/bandolier/painres/dow

nload/whatis/Cost-effect.pdf>.

Qureshi, S. et al. 2011, “Mobile cloud computing as future

for mobile applications - Implementation methods and

challenging issues”, IEEE International Conference on

Cloud Computing and Intelligence Systems (CCIS),

Conference Proceedings, pp.467 – 471.

Sanders, C. et al., 2012, “Exploring barriers to

participation and adoption of telehealth and telecare

within the Whole System Demonstrator trial: a

qualitative study”, BMC Health Services Research

2012.

Sarasohn-Kahn, J., 2010, “How Smartphones Are

Changing Health Care for Consumers and Providers”,

California Healthcare Foundation.

Steventon, A. et al., 2012, “Effect of telehealth

interventions on use of secondary care and mortality:

findings from the Whole System Demonstrator cluster

randomised trial”, BMJ2012;344:e3874 doi:

10.1136/bmj.e3874.

Steventon, A. et al., 2013, “Effect of telecare on use of

health and social care services: findings from the

Whole System Demonstrator cluster randomised trial”,

Age and Ageing 2013; 0: 1-8 doi:

10.1093/ageing/aft008.

Sundmaeker, H. et al., 2010, “Vision and Challenges for

Realising the Internet of Things”, European

Commission.

Viet, V. Q., Lee, G. & Choi, D., 2012, “Fall Detection

based on Movement and Smart Phone Technology”,

2012 IEEE-RIVF International Conference on

Computing and Communication Technologies,

Conference Proceedings, pp.1-4.

Wang, C. C. et al., 2008, “Development of a Fall

Detecting System for the Elderly Residents”, The 2nd

International Conference on Bioinformatics and

Biomedical Engineering 2008, Conference

Proceedings, pp.1359-1362.

Wei, Z. et al,. 2010, “A Residential Gateway Architecture

Based on Cloud Computing”, 2010 IEEE International

Conference on Software Engineering and Service

Sciences (ICSESS), IEEE Conference Publications,

pp.245-248.

Wu, C. Y. et al. 2009, “Introduction and Application of

Important Digital Home Standard Alliances” (in

Chinese), TLTECHNICAL Journal, Vol. 39, No.5, pp.

601-620.

Yang, Y. & Zhao, X., 2011, “Development of a Fall

Detection Algorithm based on a Tri-axial

Accelerometer”, The 4th International Conference on

Biomedical Engineering and Informatics (BMEI),

Conference Proceedings, Vol. 3, pp.1371-1374.

ZigBee Alliance, 2009, “ZigBee Wireless Sensor

Applications for Health, Wellness and Fitness”.

ZigBee Alliance, 2012, “ZigBee 2012 Specification

Features At-A-Glance”.

Zimmermann1, G. & Vanderheiden, G., 2007, “The

Universal Control Hub : An Open Platform for

Remote User Interfaces in the Digital Home”, Lecture

Notes in Computer Science, Vol. 4551, pp.1040-1049.

Cost-effectiveDesignofReal-timeHomeHealthcareTelemonitoring