Decision Support Systems 46 (2009) 606–619

Contents lists available at ScienceDirect

Decision Support Systems

j ourna l homepage: www.e lsev ie r.com/ locate /dss

Enabling ubiquitous patient monitoring: Model, decision protocols,opportunities and challenges

Sweta Sneha ⁎, Upkar VarshneyDepartment of Computer Information Systems, Georgia State University, Atlanta GA 30302, USA

⁎ Corresponding author.E-mail address: ssneha@kennesaw.edu (S. Sneha).

0167-9236/$ – see front matter. Published by Elsevier Bdoi:10.1016/j.dss.2008.11.014

a b s t r a c t

a r t i c l e i n f o

Available online 21 November 2008

Keywords:

Healthcare costs in the US afuture. Management of chrbeen widely proposed as a

Patient monitoringAd hoc wireless networksHealthcareTechnology

viable option for economizing healthcare resources, and providing efficient,quality healthcare. The process of ubiquitous patient monitoring is information intensive, the informationgenerated is not only fragmented but also spans multiple processes, artifacts, parameters, and decisioncriteria. The current study explores the complexities associated with the process of ubiquitous patient

re approximately 15% of GNP and are anticipated to reach 17% of GNP in the nearonic diseases via technology based ubiquitous patient monitoring services has

monitoring and the enabling technologies. The key contribution is a framework that captures the complexprocesses, the parameters involved, and the decision criteria for ubiquitous patient monitoring. The decisionprotocols and enabling technologies supporting the processes are detailed in the study along with theopportunities and challenges of ubiquitous patient monitoring. A conceptual model of ubiquitous patientmonitoring is developed by leveraging the proposed framework and is validated by a usage scenario. Finally,the implications of future research and contributions of the current research are discussed.

Published by Elsevier B.V.

1. Introduction

Healthcare forms an indispensable constituent of the modernsociety, representing a large percentage of Gross National Product(GNP) and sustaining a high political profile and strong public interest[5]. In the wake of the 21st century, healthcare systems around theglobe are faced with an exponential rise in expenses, heavy utilizationof services associated with a steep rise in aging population, andlimited financial as well as human resources to manage the growinghealthcare needs [13,48]. The current healthcare expenses in the USare approximately 15% of the GNP [25] and projected to reach 17% ofthe GNP by 2011 [6]. Another trend observed parallel to the risinghealthcare costs is the “graying of the globe” — the worldwidepopulation of adults over 65 years of age is on the rise and is expectedto reach 761million by 2025 [46,63]. Fig. 1 depicts the changing globaldemographics, the resultant increase in the number of aging patients,and the corresponding strain on both the human as well as financialresources of the healthcare sector.

Multiple studies in the past have noted the prevalence of chronicdiseases in the aging population — seven of the most prevalentchronic illnesses in the US (and their associated in-patient expenses)include: coronary artery diseases ($25.6 billion), heart failure($15.2 billion), chronic obstructive pulmonary diseases ($6.2 billion),mental health disorders ($3.9 billion), diabetes ($3.8 billion), hyper-tension ($3.2 billion) and asthma ($1 billion) [6]. Medicare's high-risk

.V.

patients, approximately 8 million currently, with five or more chronicdiseases account for approximately 78% of all health care spending —

well over a trillion dollar per year and/or over two-thirds of Medicare'sannual spending [3,6,17]. Many healthcare experts agree that currentMedicare expense patterns are a reflection of chronic illnesses managedunsuccessfully [6]. A large percentage of chronic diseases deteriorate tothe point where a crisis is reached resulting in unnecessary long-termhospitalization at massive cost to the healthcare sector. A criticalinference drawn from epidemiological data and past studies is thatpreventing occurrences of acute episodes holds the key to providingquality healthcare, reducing incidences of prolonged hospitalizationsand resultant healthcare expenses [43]. In order to reduce preventableacute episodes from occurring, it is critical to focus on preclusion ofcrisis/complications, proactive management of chronic illnesses, andtimely detection of anomalies such that patients can lead a normal,healthy lifestyle outside of the hospitals.

Innovative strategies are needed to tackle the spiraling healthcareexpenses and to cater to the healthcare needs of an aging populationin addition to sustaining the trend towards an independent lifestylefocusing on personalized non-hospital based care [46,60]. Onestrategy is deployment of a large number of trained healthcareprofessionals to handle the current healthcare scenario involvingchronic illnesses. However, there are two key constraints associatedwith heavy utilization of human resources towards timely detection ofanomalies, prevention of complications, disease management, educa-tion/guidance with respect to medications, exercise, and diet (withinthe context of chronic illnesses): (1) healthcare professionals arelimited and over-worked; (2) human resources constitute the most

Fig. 1. Changing global demographics and resultant healthcare needs.

607S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

expensive variable in the healthcare sector. Thus heavy utilization ofhuman resources will not only increase the cognitive overload of thehealthcare professionals, but will also increase healthcare costs.

This research explores ubiquitous monitoring of chronic patients as apotential solution that seeks to: (1) leverage ubiquitous biomedicalsensing, computing, and communication technologies to complementand assist healthcare professionals in efficiently managing chronicillnesses 24×7, (2) reduce incidences of unnecessary hospitalizationsdue to undetected complications, (3) provide timely detection ofanomalies before it snowballs into a crisis, and (4) provide pertinentmedical attention utilizing the expertise of the healthcare profes-sionals for handling anomalies “just-in-time” as and when neededwithout time and/or location dependency [5,14]. The populationtargeted for ubiquitous monitoring consists of Medicare's high-riskpatients suffering from multiple chronic illnesses and incurring thelion's share (over a trillion dollar) in annual healthcare expenses.Ubiquitous patient monitoring does not intend to replace existinghealthcare systems, practices, and personnel instead it seeks to assistand complement the functioning of the existing healthcare systemstowards efficient utilization of resources and reducing unnecessary ex-penses incurred due to poorly managed chronic illnesses.

The market for patient monitoring services is a multi-billion dollarindustry [61] that holds the potential to reducing unnecessaryhealthcare costs while providing quality care to an aging populace.Hence there is an imminent need to fully explore the domain ofubiquitous patient monitoring with respect to critical factors andenabling technologies. However, there is a dearth of research focusingon specific understanding of the role of technology based ubiquitousmonitoring of patients in improving the practice and delivery ofhealthcare. Thus the goal of the current research is to:

• Explore the paradigm of ubiquitous patient monitoring within thecontext of chronic illnesses focusing specifically on successful diseasemanagement i.e., preventing incidences of unnecessarycomplications,prolonged hospitalizations, and corresponding expenses;

• Provide clear guidelines focusing on the process of patientmonitoring, key parameters involved, decision logic, and an under-standing of the technology enabling efficient and effective ubiqui-tous monitoring of chronic illnesses.

Active patient involvementwithin theubiquitous patientmonitoringparadigm seeks to reduce the mental as well as physical strain of the

healthcareprofessionals, increasecompliancetotreatment, reduceunex-pected hospitalization expenses and promote a better, healthier lifestyleof patients outside the hospital. The sole purpose of this research is toexplore the utility of ubiquitous patient monitoring as a means ofeffective and efficient utilization of limited healthcare resources towardsan efficient healthcare delivery within the context of chronic illnessesand corresponding expenses. The next section articulates the concept ofubiquitous patient monitoring, the associated requirements/challenges,and the enabling technologies.

2. Ubiquitous patient monitoring

Ubiquitous Patient Monitoring is a concept that has its roots in thevision of MarkWeiser— the father of ubiquitous computing. His visionof ubiquitous computing is captured beautifully in the followingquote: “The most profound technologies are those that disappear.They weave themselves into the fabric of everyday life until they areindistinguishable from it” [62]. Healthcare seems to be themost fertileground for ubiquitous computing applications since there is no otherdomain where the importance of making correct decision based onobtaining the right information at the right time is more critical[39,62].

The promise of ubiquitous patient monitoring is an environmentconstituted flawlessly by enabling technologies that promote contin-uous, reliable monitoring of patient specific medical informationwithout any dependence on time and location such that promptmedical intervention is provided as and when needed. The informationobtained is analyzed via ubiquitous computing technologies for timelydetection of anomalies and promoting compliance. Consequently,ubiquitous monitoring solutions, both short term and long term atpatients' homes, nursinghomes, and hospitals, are increasingly seen as aviable option for disease management, reduction in episodes ofpreventable hospitalizations, corresponding expenses, and provisioningof healthcare services “just in time” as and when needed [41].

The definition of ubiquitous patient monitoring involves two per-spectives, one being the domain of application of the technologiesenabling ubiquitous computing and the other being the concept thatintegrates healthcare more seamlessly to our everyday life [27].Ubiquitous patientmonitoring is notmerely a technological innovation;it involves a paradigm shift in healthcare practice, delivery and view.This paradigmshift implies technical applications of consumer-operated

608 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

interoperable standard technologies for health and wellness leveragingthe advances in the three classes of technology — ubiquitous sensing/monitoring, computing, and communication such as: PDA (PersonalDigital Assistant), mobile phones, and communication networks. At thelevel of the healthcare organization, ubiquitous healthcare implies achange from physician-centric systems to patient-centric operationalmodels [28].

Ubiquitous patient monitoring services seek to assist the physi-cians in managing chronic illnesses and not replace their expertise.While ubiquitous patient monitoring services focus on promotingcompliance and detection of anomalies, the focus of the physicians ison provision of pertinentmedical attention as andwhen an anomaly isdetected without any delays. The core responsibilities of thehealthcare professionals will shift frommonitoring patient conditions

Fig. 2. A multi-dimensional model of ubiqu

arising frommismanagement of diseases and undetected anomalies toprovision of medical expertise. Ubiquitous technology will step in totake care of the standard, repetitive tasks of monitoring vital signs andpromoting compliance with medical advice, performing analysis, andrequesting medical assistance only when required.

A multi-dimensional model of ubiquitous patient monitoring de-picting the enabling technologies, the benefits to the healthcare sector,and the various facets of healthcare services that can potentially besupported in such an environment is depicted in Fig. 2.

2.1. Ubiquitous patient monitoring — requirements/challenges

The requirements of patient monitoring are not only diverse sup-porting indoor and outdoor, as well as stationary and mobile patients

itous patient monitoring environment.

609S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

but are also complex, and involve multiple parameters such as:duration of monitoring, frequency of data collection and transmission,amount of data transmitted, nature of monitoring such as: alert,periodic or continuous. The following overview of the requirements ofpatient monitoring shows the complexity, diversity, and somewhatcontradictory nature of the requirements. Fig. 3 graphically depicts aconceptual classification of the diversity and complexity of patientmonitoring requirements.

2.1.1. Monitoring and transmissionPatient monitoring can be continuous, alert driven (on the

detection of an abnormal event) or periodic (at fixed times in a day).Continuous monitoring and transmission can provide real-time datafor analysis and storage but can also lead to an information overload ofhealthcare professionals and significant network traffic. Periodicmonitoring and transmission sacrifices the real-time aspect for adecrease in network traffic and information overload and is suitablefor patients under routine supervision. In alert driven monitoring, thepatients are continuouslymonitored and the acquired data is analyzedfor detection of anomalies. Only when an anomaly is detected andmedical attentionwarranted, is amessage transmitted over a network.This type of monitoring could result in a time lag in performinganalysis and sending alert messages, however, a reduced informationload on healthcare professionals who would otherwise be faced withthe task of analyzing the incoming data and network traffic make it apossible choice in some instances of patient monitoring.

2.1.2. Reliability of message deliveryDue to the potentially life threatening situations, reliability of

message delivery to healthcare professionals is the most criticalrequirement of patient monitoring. Routine transmission of signalsfrom a patient can tolerate low reliability while emergency messageshave the highest reliability requirement. Thus different monitoringmessages can be prioritized based on the reliability requirements. Thefactors impacting reliability in infrastructure and ad hoc wirelessnetworks based solutions include: network coverage, presence of deadspots, device range, available power, bit rate, routing protocol, failure(s)in the network or device, and un-cooperative behavior of other devices.

2.1.3. Reasonable time in message deliveryAny delay in message delivery can have fatal consequences. The

priority of transmitted message (emergency or routine) can be used to

Fig. 3. A conceptual framework classifyin

determine the routingofmessages bya network to reduce delays,whichmay be substantially impacted by frequency of monitoring, size ofmessage transmitted, bit rate, and the number of monitored patients.The objective is to haveminimum delay in end to endmessage delivery.

2.1.4. Power conservationPower management of the low powered monitoring devices with

diversity in the range of functionality and computing capabilities usedfor signal transmission poses a challenge in providing reliable patientmonitoring solutions. The battery (power source for most wearabledevices) typically forms the heaviest component; hence there is atradeoff between carrying a heavy device and the frequency ofrecharging the battery. The critical factors impacting utilization ofpower include the frequency and size of transmitted messages androuting schemes employed for message transmission.

2.1.5. Support for mobile patientsPatient monitoring solutions should be able to support mobile

patients indoor and outdoor. Patient mobility results in patientsmoving in and out of network coverage (in an infrastructure basednetwork), thus negatively impacting reliability in patient monitoring.Hence the challenge is to design dependable networking support formonitoring both mobile and stationary patients.

2.1.6. ScalabilityThe patient monitoring network must scale well in terms of the

number of monitored patients that can be reliably supported. Thefactors influencing scalability are: bit rate, frequency of monitoringand transmission, and the amount of information transmitted perpatient.

2.1.7. Manageable cognitive load for healthcare professionalsAnalyzing continuous streams of data frommonitored patients and

making relevant diagnosis can be an arduous task for the healthcareprofessionals. It can also impact network traffic and scalability of thepatient monitoring systems. The solution promised by the ubiquitouspatient monitoring environment is utilization of the computationalcapabilities of the monitoring devices for intelligent analysis anddeductions, and alerting the healthcare professionals only when ananomaly is detected. The collected information can be accessed by thehealthcare professionals (if required) to make informed decisions andclosely strategize diagnosis, treatments/outcomes, and patient

g patient monitoring requirements.

Table 1Requirements, decisions, and technology supporting patient monitoring

Requirement Decision criteria and enabling technologies

Monitoring What to monitor — BP, ECG, body temperature, bloodOxygen levelWhen to monitor — periodically (what time interval),continuouslyHow to monitor — via sensors, physically separatedevices ex: thermometer, BP monitor.

Analysis Analyze the collected information before transmission?What to analyze — the vital signs, the compliance factorsHow to analyze — based on stored thresholds,standardized dataHow/what to transmit — which network, patient'slocation, current reading

Transmission What to transmit— complete recorded data, differentialdata, personal informationHow to transmit—WLANs, cellular PCS/GSM, or ad hocwireless networksWhen to transmit — periodically, continuously, alertbased

Reliability Reliability in sensing, analyzing and transmitting patientinformation via built in redundancyVarying reliability levels for transmitted messages?- High (emergency), medium (routine), low(reminders)- Reliability in communication enabled by utilizingmultiple communication networks

Delays Prioritized transmission. Low— emergency transmission,medium— routine transmission

Power conservation Mobile computing and communication devices withpower conservation methods such as sleep cycles,emergency power reserve, solar energy, innovation inconserving/charging devices.

Support for mobility Patients mobile indoors/outdoors?Variation in mobility pattern — high/low speed,children/adults

Cognitive load onhealthcare professional?

The analysis of the plethora of information collectedpatients monitored electronically can be an arduous task.However, the cognitive load on the healthcareprofessional should be low by automating the analysis ofthe obtained medical information and involving thehealthcare professionals only when required.

Security, privacy, andconfidentiality

High levels of security, privacy and confidentiality.Technologies with built in secure authenticationprocesses, secure fool-proof communication resources,biomedical measurement/sensing devices, and patientspecific intelligent devices

Proactivity and transparency Varying levels of proactivity and transparency? Low -N moreinput, high -N low inputIntelligent environment with the potential of learning andreacting to anomalous conditions.

Invisibility Pervasive presence/disappearance of computing/mobilecommunication resources, and sensing devices in everydayobjects such as smart shirts

Context awareness High level of context awareness in order to correctly analyzethe collected information

610 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

education towards prevention of crisis and reducing episodes ofpreventable hospitalizations.

2.1.8. Confidentiality, security, and privacyAs healthcare information is being transmitted over wireless

networks, efforts should be made to keep it confidential and private.Privacy entails the right of a user to control the collection anddissemination of personal information and security is the protectionof user's information from unauthorized users. Privacy and securityare one of the key challenges towards large scale adoption anddiffusion of ubiquitous computing empowered by wearable devices[24,60]. But depending on the context privacy and security can be onthe opposite ends of the spectrum where increasing security couldimply lowering privacy and vice versa. This is expected to be one of themost critical requirements for healthcare administrators and govern-ment regulators.

2.1.9. InvisibilityOne of the most promising concepts of ubiquitous healthcare is

providing healthcare services pertaining to monitoring, treatment(reminders to take medication), prevention and management ofdiseases in a manner that causes minimal distraction and isunobtrusive in nature. The devices/computers promoting ubiqui-tous monitoring should disappear in the background, removingdistractions by computers such that the patients can continue withtheir daily activities without being obstructed by the computers[13,39,62].

2.1.10. Proactivity and transparencyProactivity and transparency refers to the intelligence in the

ubiquitous healthcare environment which will be able to sense thecurrent intent of the patient and proactively take certain actions onbehalf of the patient. For instance: if the device senses that thepatient's ECG has gone beyond a pre specified threshold then theintelligence in the device may want to alert the doctor of the situationand schedule an appointment as soon as possible along withinforming the patient of the appointment. But this proactive actionshould be transparent to the patient as far as possible. The proactivityshouldn't become a source of annoyance.

2.1.11. Context awarenessContext awareness is in synch with proactivity since a patient

monitoring environment cannot be proactive in assisting a user indecision making unless the right context information is available. Forinstance, if the patient's hear rate has gone up then the device must beaware of the context where and when the heart rate was high. If thepatient was watching an exciting football game which caused theheart rate to go up while the vital signs from other sensors werewithin the normal range then the device should be able to detect thecontext and not send alerts. Context information can be recoveredfrom sensors and would essentially contain information about who,what, when and where.

Table 1 presents the requirements/challenges associated withubiquitous patient monitoring along with the various decision factorswhich characterize each requirement.

The technologies supporting ubiquitous monitoring of patientsexists and is much more sophisticated today than their precursors adecade ago; however the utilization of such technologies forpromoting ubiquitous monitoring of patients doesn't come withoutits own challenges. Despite of the limitations, there are massiveopportunities and promises associated with ubiquitous patientmonitoring. Some of the key limitations associated with the usage/implementation:

• Feasibility analysis with respect to cost/benefit — healthcare sectoris running in red and an analysis of the cost and benefits in terms of

financial feasibility, return on investment, and quality of serviceissues focusing on reduction in hospitalizations and incidences ofcrisis are critical issues that are yet to be fully explored;

• Acceptance/adoption by the healthcare professionals and targetpopulation — healthcare sector has traditionally been slower inadopting technological innovations in the practice and delivery ofhealthcare than other major sectors such as banking and automotive.EMR (Electronic Medical Records), although endorsed by the govern-ment and supported by several technology giants, has met withmassive resistance in adoption and usage;

• Actual use of innovative technology based ubiquitous patient mon-itoring and related benefits are additional important concerns thathasn't been assessed/quantified by any prior studies to the best of ourknowledge.

611S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

3. Evolution of patient monitoring solutions

The current advancement in wireless communication technologyand patient worn devices for medical telemetry has given a boost tomonitoring solutions for patients inside and outside the hospitalpremises. It is now possible to record and transmit digitized vitalssigns in the form of signals from a patient device to the computer orhand-held PDA of healthcare professional instantly, hence reducing thetime taken for evaluation and treatment [14,16,26,35]. In the recent past,research and development community has given considerable attentionto understanding and developing viable patientmonitoring applicationsand research prototypes for accurately monitoring patients' vital signsand timely detection of anomalies [41]. Several companies, includingGeneral Electric, Hewlett Packard, Honeywell, and Intel, have teamed upin the Center for Aging Services Technologies (CAST) inWashington D.C.,established in 2002, to encourage collaborative aging-related technologydevelopment and to advocate wireless, remote monitoring of patients,specifically the aging populace that incurs the largest percentage ofhealthcare expenses.

First generation monitoring services, involve collection and transmis-sion of data within a hospital using wireless LANs such as Micropaq thattransmits multi-parametric information via Wireless LANs [57] andLifeSync [56], which uses short rangewireless signals such as Bluetooth tomove information within a hospital infrastructure. Next generationmonitors, such as Medtronic [34], allow patients to live in their homeswhile required data is collected and transmitted at predefined time— endof day or week. Motiva provides disease management and increasedquality of life to the patients via a secure, personalized healthcarecommunication platform that connects chronic patients at home to theirhealthcare providers through their TV sets and cable systems for IP access[59]. Some monitors provide short term monitoring for 10–14 days withthe intention to diagnose a patient's condition tomatch itwith the correcttreatment such as: CardioNet [52], and Biotronik [51]. A system for real-timemonitoringofpatients in thehomeenvironment ispresentedby [26],asthma in homemonitoring via a video monitor and a secure website foruploading the relevant information is presented by [9], and unobtrusivewellness monitoring of elders via sensors by IST Vivago Wristcare withautomatic alarm triggering capacity and communication over telephonelines [38]. Recently, prototypes and research have focused on providingmonitoring solutions to mobile patients by exploring continuouscollection and transmission of vital signs via infrastructure-basedwirelessnetwork (Wireless LAN, Cellular PCS, and Satellites) [31,35,41]. Work onother related issues inpatientmonitoring include “smarthealthwearable”research [32], interference for telemetry devices [12], PDA as a mobilegateway [22], long-term health monitoring by wearable devices [1,45],and, smart shirt based health monitoring [58], a wearable stethoscope[29]. Clothing-embedded transducers for ECG, heart rate variability, andacoustical data andwireless transmission to a central server are proposedin [22]. A requirementmodel for delivering alertmessages is presented in[23]. A design approach for data compression for amobile tele-cardiologymodel is presented in [19], where a significant compression ratio andreduction in transmission time over GSM network was achieved.Strategies for efficiently working with digital medical images is describedin [42]. Personal health monitors based on wireless body area network(BAN) of intelligent sensors are proposed for stress monitoring [21]. AlertBased continuous monitoring of Parkinson patients by intelligentwearable devices is described in [46]. Monitoring of Alzheimer patientsin geriatric residents via an autonomous intelligent agent and effective-ness of remote asthma monitoring is described respectively in [8,11].Numerous additional means of improving the delivery and quality ofhealthcare to patients via improvedmedical decisionmaking is describedin [4,15,30,49]. Empowered by technology promoting ubiquitous comput-ing, there has been some research and development efforts for intelligentmonitoring of patients in context aware, secure, smart environment forassisting elders to living independently, examples include Gator TechSmart House [55], Aware Home Project [50], and Elite Care [44].

Although the prior body of research with respect to patientmonitoring is noteworthy and has made some outstanding contribu-tions, the primary limitations include the following: (1) the solutionsproposed thus far have not delved in the realm of ubiquitousmonitoring — thus being inherently constrained in terms of patientmobility, continuous monitoring and timely detection of anomalies,(2) majority of the solutions do not fully utilize the mobile,computational platform to perform analysis of biometric data priorto transmitting thus resulting in high network traffic, higherbandwidth requirement, and cognitive overload of the healthcareprofessional with the daunting job of analyzing streams of biometricdata and detecting anomalous conditions. There is vast disparity in thedegree of reliability and effectiveness of remote patient monitoringservices as provided by commercially available portable monitors andresearch prototypes in the context of timely detection of anomaliesand provision of medical attention. One potential source of suchdisparity can be attributed to the lack of understanding of the complexprocess, the associated parameters, and the decision protocolsassociated with patient monitoring solutions. The current studyparticularly adds value to the rich body of research by developing aconcise framework that supports comprehensive, ubiquitous mon-itoring of chronic illnesses with the goal to: (1) provide timelydetection of anomalies and prevent incidences of acute episodes thusreducing avoidable hospitalizations and corresponding expenses, (2)leverage the computational and processing capabilities of thetechnologies enabling ubiquitous monitoring such that the biometricdata is transmitted to the healthcare professional only when ananomaly is detected and medical intervention is warranted; therebyleading to decreased cognitive overload and efficient utilization ofhuman resources, reduced network traffic and bandwidth require-ment. Thus the objectives of the current research with respect to theparadigm of ubiquitous patient monitoring and its role in assisting thehealthcare sector in providing quality healthcare to chronic illnessesare as follows:

• Articulating the concept/definition of ubiquitous patient monitoringand assessing and quantifying the corresponding complex require-ments/challenges (section 2).

• Developing a concise conceptual framework capturing the (a) keyelements and information flow associated with the process of ubiq-uitouspatientmonitoring, and (b) the roleof theprimarystakeholders—patients/physicians within the framework (section 4).

• Modeling and detailing the fundamental elements constituting thecomplex process of technology based ubiquitous patient monitoringprocess, specific parameters, decision protocols, information ana-lyzed/generated, and enabling technology corresponding to theprocesses (section 4).

• Developing a conceptual model of a ubiquitous patient monitoringsolution as a proof of concept (section 5).

• Designing a usage scenario to explore the utility of ubiquitousmonitoring of patients as a man–machine problem solving systemutilizing technology for providing support to the structuredprocesses of patient monitoring while leveraging the tacit expertiseof healthcare professionals to provide support to the unstructuredprocesses (section 5).

• Discussing specific guidelines for further research, limitations, andchallenges of ubiquitous patient monitoring (section 6).

4. Proposed framework for ubiquitous patient monitoring

The underlying covenant of technology enabled patient monitor-ing services is that prompt medical attention will be provided to thepatients “just-in-time” as and when required without any constraintsbased on time and location. Comprehensive, reliable patient monitor-ing outside the hospitals will not only reduce the healthcare expensesassociated with hospitalizations but will also allow precious human

Fig. 4. A framework characterizing the processes of patient monitoring.

Fig. 5. Decision protocol for the process of sensing.

612 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

resources to be allocated to the healthcare needs of other patients[7,10,33]. An effective patient monitoring solution can be leveraged asa decision support tool where the structured/routine tasks in thepatient monitoring process is tackled by technology while theunstructured tasks are tackled by the healthcare professionals. Thiswill not only economize the human resources of the healthcares sectorbut will also enable improved healthcare of the aging population.

Ubiquitous patient monitoring system constitutes a man–machineproblem solving system, utilizing ubiquitous computing, communicationandsensing technologies forproviding support to thestructuredprocessesof patient monitoring and the tacit/specialized knowledge of a healthcareprofessional to provide support to the unstructured processes. Thestructured process in the process of patient monitoring involves: (1) sen-sing/obtaining specific vital signs such as ECG, oxygen saturation level,body temperature, blood pressure, and heart rate, and patient parametersand information such as skin breakdowns, abnormal gait and balance,motoractivityandagitation, current location,weight, cigarette smoke, andthe amount of moisture in clothes and/or levels of physician specifiedchemicals as in cancer treatment, (2) analysis of recorded vital signs, and(3) transmitting thedatavia a communicationnetwork (wireless orwired).

The proposed framework consists of the following: (1) a processmodel depicting the process of patient monitoring, and (2) decisionprotocols that support each process considering the informationgenerated/analyzed, specific parameters that impact each step, anddecision criterion for accomplishing structured tasks. The framework(shown in Fig. 4) is then discussed in detail in the context of eachprocess, decision steps, the enabling technologies, the requirements/challenges, the decision protocols, and the decision criteria/para-meters impacting each process.

4.1. Sensing

Sensing is the basic process that entails collecting data regardingspecific conditions including but not limited to the following: vital

signs such as ECG, oxygen saturation level, body temperature, bloodpressure, and heart rate, and patient parameters and information suchas skin breakdowns, abnormal gait and balance, motor activity andagitation, current location, weight, cigarette smoke, the amount ofmoisture in clothes and/or levels of physician specified chemicals as incancer treatment, and, routine intake of prescriptions medications asspecified. Continuous, comprehensive sensing of patient specificconditions is enabled by the devices that can collect multi-parametricvital signs of patients irrespective of time and place, and the sensors/motes that check specific conditions such as sudden increase in

Fig. 7. Decision protocol supporting the process of transmission.

613S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

weight and missing prescriptions. Wearable biomedical sensors andbiomedical clothing and other monitoring devices have made possiblethe collection and analysis of physiological patient data while thepatient is mobile [1]. Inexpensive wireless heart-rate monitors havebeen available for consumers for several years ex: such as CardioNet[52], which collects ECG data for detection of arrhythmia. The widespread influx of handheld andwearable computers, miniaturization ofprocessors, development of intelligent textiles (Smart Shirt [58]),smart papers, natural communication of the user via intelligentinterfaces capable of interpreting speech and gesture are bringing uscloser to the vision of ubiquitous monitoring. The technology tosupport ubiquitous sensing already exists, however the biggest barrierlies in seamless integration and interoperability of the technologysuch that the patient can lead a normal life with minimal distractionfrom the technology sensing all of the health related conditions. Fig. 5shows the decision factors considered in defining the processmonitoring. The decisions will typically be made under the doctor'srecommendation by the patient. The key consideration made in theprocess of sensing is the frequency and reliability in data collection.

4.2. Analyzing

Analyzing is the process of evaluating the data collected viasensing in order to detect the presence of any anomalies and takecorrective actions as prescribed such as: transmit alert messages to ahealthcare professional regarding the detection of anomalies, and call911 (see Fig. 6). The process of analyzing the collected data beforetransmitting not only reduces the cognitive overload of the healthcareprofessionals but also increases the scalability and throughput of thecommunication channel, however there is a little bit of time lost inanalyzing before message transmission. The key considerations madein the process of analyzing the collected data are: to maximize thefault-tolerance of the process so that the anomalies can be detected inan error-free manner, and power conservation of the monitoring/analyzing devices. The technology enabling this is ubiquitouscomputing which seeks to bridge the gap between the virtual andphysical world by incorporating computing power (microprocessors)and sensing (sensors) into anything, including not only conventionalcomputers, personal digital assistants (PDAs), mobile phones, printers,but also everyday objects like white goods, toys, plates, cups, glasses,houses, furniture, or even paint (“smart dust”).

Fig. 6. Decision protocol for the process of analysis.

4.3. Transmitting

Ubiquitous communication technologies supporting reliable trans-mission of signals between sensors, patient worn computing devices,and healthcare are critical to the success of ubiquitous patient moni-toring (see Fig. 7). Ubiquitous communication allows continuousaccess to data and communication power between people andartifacts with computing and communication capabilities. Mobileand wireless communication technologies and ad-hoc networking aresome of the key technologies with respect to ubiquitous communica-tion [27,48,49]. The key consideration made in the process oftransmission is to support patient mobility while maximizingreliability of transmission and conserving power of the transmittingdevice.

The increase in processing power and communication bandwidthalong with the corresponding decrease in the cost of processing andcommunication,with respect to both hardware and software and powerconsumption have been the fundamental trend in information andcommunication technologies (ICT) [13,28,48]. This trend is making itboth technically and economically possible to integrate processingpower and communication capacity to more simple and inexpensivedevices andobjects.Withmore than2billionmobile devicesworldwide,it is evident that the PDAs and mobile devices are slowly becoming anintegral part of our lives [48]. The plethora of mobile devicessurrounding us with computing and processing capacity offer a mobilewireless platform for running healthcare applications, a user interfaceand data logging potential for health sensors and monitoring devicesand a gateway to connect local devices collecting health relatedinformation to global services such as a hospital databases [41,48].RFID (radio frequency identification technology) allows simple wirelesscommunication with nearby objects to obtain information such asproduct code, URL or sensor reading. The RFID tags are becomeinexpensive to be produced in mass scale and do not require batteryfor operation since they use backscattering in communication.

Transmission of data for patient monitoring has so far beensupported by either wired or wireless infrastructure based network.Periodic monitoring in the home environment has been supported viatelevision sets, video monitoring, and two-way broadband connectionvia cable modem and telephone lines. The wired communication isreliable but doesn't support patient mobility and continuous

Fig. 8. Multi-hop signal transmission in ad hoc wireless network.

614 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

monitoring. Bluetooth has been utilized for short range wirelesscommunication between the monitoring device and WLANs. Satellitebased communication has been used for GPS (Global PositioningSystem) for outdoor purposes and indoor location tracking is typicallysupported via RFID (Radio Frequency Identification). WLANs andcellular networks have primarily been used to support continuouspatient monitoring along with patient mobility. The transmissionchannels to transmit the required information include:

- Short range transmission between the sensing and analyzingdevices via Bluetooth.

- Infrastructure oriented wireless network such as WLANs, CellularPCS/GSM, and Satellite Based Networks. The spotty coverage ofexisting infrastructure oriented wireless networks (such ascellular networks and wireless LANs) due to time and locationdependent channel quality and signal attenuation results indead spots and can consequently lead to unpredictable qualityand reliability of monitoring solutions [48]. The exclusivedependency on infrastructure-oriented wireless networks forpatient monitoring has additional challenges including: (a) shortrange and limited power capabilities of most patient devices,(b) lack of interoperability among multiple wireless LANs,(c) considerable interference in ISM bands from multiple sources,(d) varying capacity of infrastructure-oriented wireless networks,and, (e) lack of application-specific priority for transmission ofemergency signals. These restrictions and requirements combinedwith a lack of comprehensive coverage of infrastructure-orientedwireless networksnegativelyaffects thequalityof patientmonitoring,

Fig. 9. Conceptual model of a propo

greatly limits themobility of patients, and can potentially lead to fatalconsequences.

- Ad Hoc Wireless Networks formed among patient monitoringdevices can complement the coverage of infrastructure orientednetworks in the areas with limited/no network coverage frominfrastructure-oriented networks thereby leading to improvednetworking support for patient monitoring solutions and conse-quently enhancing the quality and dependability of patientmonitoring solutions. There are numerous opportunities andchallenges associated with leveraging mobile ad hoc networkto support dependable patient monitoring solutions [47]. Amobile ad hoc network consists of a collection of geographicallydistributed wireless devices or nodes that can dynamicallyform a network without a pre-defined infrastructure and com-municate with one another over a wireless medium [20,36,37](Fig. 8).

The advantages associated with ad hoc networks are ease andspeed of deployment since it can function dynamically without anyinfrastructure, robustness, flexibility in terms of place of deploymentand number of users and lastly the inherent support for mobilitywhich forms the cornerstone of deploying such networks in health-care since the patients and doctors are both mobile [37,40]. Thechallenges are related to the variations in the mobility pattern andmobility characteristics of the nodes; asymmetric capabilities of thenodes with respect to transmission range, battery power, processingcapacity and mobility; diversity in traffic characteristics such as bitrate; timeliness constraints, and reliability requirements. The routefrom the source to the destination typically involves multiple hopsand the route is susceptible to changes due to mobility in nodes.Recent advent in personal digital assistants and plethora of mobilepatient monitoring devices that are used in transmission of vital signsover short range, have brought to the fore ground the possibility offorming ad hoc networks among patients' devices which can monitorand transmit vital signs.

5. Conceptual model — a proof of concept

Although the technology to support ubiquitous patient monitoringexists, the key question is how to create an integrated architecture toachieve the vision. A critical factor leading to increased healthcareexpenses is hospitalization for long term care and monitoring. Hence

sed patient monitoring system.

Table 2Monitors and Sensors in the Proposed Conceptual Model

Monitors andsensors

Objective/functionality

ECG monitor Measures ECG of patients. The portable ECG is carried by the patient 24 h a day. The ECG device has two electrodes that are attached to the patient andprovide a single channel ECG Signal (Nussbaum et al., 2002)

Blood sugar monitor Monitors blood sugar level at predetermined times as deemed suitable by a doctor.Asthma monitor Monitors peak air flow level at predetermined times as deemed suitable by a doctor.Medication sensor Located on the prescriptions (Rx) bottles. Monitors the intake of Rx as advised. If violation detected then relevant signal is sent to the Medication Agent in

the patient's PDA.Weight sensor Located on the weighing machine. Senses any critical change in the patient's weight based on pre-stored patient's weight. If a critical change in weight

observed then signal is sent to the Weight Agent in the patient's PDASleep sensor Located on the bed. Senses sleep patterns. If sleeplessness or too much sleep is sensed over a 24 h period, then the signal is sent to the sleep agent

615S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

shifting the site of continuous monitoring and care from the hospitalto the patient's home can potentially reduce healthcare cost [2,10]. Thenext section presents a proof-of-concept for ubiquitous patientmonitoring as a detailed conceptual model. The model providesdetails with respect to the various components, the functioning, andthe decision criteria considered by each component.

5.1. Conceptual model

The proposed framework is utilized in developing the conceptualmodel. Fig. 9 represents the conceptual model of a ubiquitous patientmonitoring environment equipped with sensors and biometricdevices. Each process is broken down into decision segments asoutlined in the framework. The development of the conceptual model,the underlying processes, the corresponding parameters, and thedecision criteria are discussed in the next few paragraphs.

5.1.1. What population segment is served by the proposed model?The conceptual model is designed to meet the healthcare needs of

the Medicare's high-risk patients (approximately 8 million) with fiveor more chronic conditions accounting for over two-thirds ofMedicare's annual spending [3,6]. A large percentage of chronicdiseases deteriorate to the point where a crisis is reached resulting inlong term hospitalization and monitoring of patients at huge costs tothe healthcare sector. The proposed study focuses on the followingchronic illnesses in the US (and their associated in-patient expenses)include: heart failure ($15.2 billion), diabetes ($3.8 billion), hyperten-

Table 3Intelligent agents performing specific tasks

Intelligent agents and corresponding tasks

Bluetooth agentResides in the PDA, retrieves readings from the wireless ECG monitor, blood sugarmonitor, asthma monitor, medication sensor, weight sensor, and sleep sensor.

Location agentLocation agent uses GPS and/or other location tracking systems (WLANs, RFID) forlocation of patients.

Patient data agentStores information about the patient such as: ID, name, address, date of birth (DOB),medical contacts information, scheduling information, blood group, and prescriptiondrugs.

Medical info. agentUpdates the patient's EMR (electronicmedical record)with the following informationreceived from the alarm agent: the anomaly detected, the date/time of the event, theID of the physician receiving/handling the alarm.

Alarm agentReceives alarm messages sent by the PDA of the patients. Informs a physician of theevent who takes the requisite action based on the analysis sent in the message andpast patient history accessed via the EMR. Sends back acknowledgement to therespective agent confirming that the alert has been received and is being managed.

sion ($3.2 billion) accounting for a total of $22.2 billion in totalMedicare expenses [3,6].

5.1.2. What is monitored and how?The proposed model monitors a patient's vital signs and specific

health and compliance conditions. The specific vital signs that wepropose to monitor and the corresponding means of monitoring arepresented in Table 2. Additionally, sensors with computing andcommunication capabilities will be deployed at specific locations inpatient's home to monitor a specific compliance task (refer to Table 2).The readings from the monitors and sensors are communicated withthe intelligent agents located in the patient's PDA. The frequency ofmonitoring the vital signs and specific conditions can be adjusted by ahealthcare professional as deemed suitable.

5.1.3. What is analyzed and how is the analysis done?We propose a PDA (Personal Digital Assistant) housed with

intelligent agents to perform the task of analyzing the monitoredconditions. The use of PDA as a component of the proposed systemstems from its ease of use, ubiquitous access to data, and the inherentsupport for mobility. The use of intelligent agents with clear goals andrelevant knowledge, as proposed in the current research, has thecapability of assisting healthcare professionals in continuous patientmonitoring in the form of ongoing analysis and diagnosis of largeamount of complex data and alerting the health center in case of ananomaly via multi agent communication [41]. This has the benefit ofnot only reducing the cognitive overload of health care professionalsbut also promoting timely intervention of health care structure as andwhen required. The intelligent agents use an ontology that cate-gorizes different alerts, which correspond to violations of boundaryconditions as specified by the threshold ECG recordings. The ontologyis represented using DAML+OIL [53], which is based on DescriptionLogics [54]. These agents carry out a large chunk of analysis of the ECGin the PDA and send alerts only if an anomaly is detected therebyreducing traffic on the wireless networks, making the system morecost-efficient and involving the health care structure without delaybut only when needed. Some of the major telecommunicationcompanies such as Motorola, Fujitsu and BT are taking great initiativesin providing accessibility to agent based services via mobile devicessuch as mobile phones, PDA or portable PCs [41]. Table 3 presentsfunctionalities of the intelligent agents which are responsible forcarrying out certain pre-defined tasks. Table 4 presents the protocolsand functionalities of intelligent agents tasked with the job ofanalysis.

5.1.4. Configuration of the patient's monitoring device — how is theanalysis done?

Prior to the first use, the Intelligent Agents tasked with analyzing aspecific patient condition is configured by a physician based on thecurrent/past conditions of the patient, thus providing individualpatient-centered care. The intelligent agents analyze the data collected

Table 4Intelligent agents — corresponding protocols and functionality

Intelligent agents and corresponding decision protocols

ECG AgentThe ECG Agent receives ECG data via Bluetooth, checks the ECG data for any anomalousconditions. If an anomalous condition is detected then it sends a request for intensivemonitoring to the ECG monitor. The readings received during intensive monitoring areanalyzed and if all of them meet the violation criteria as defined by the alert conditionsdefined in the ontology then an alarm is triggered. The alarm triggered is sent to theAlarm Agent at the E-Health Center.

IF ECG[E]b=ECG−10 OR IF ECG[E]N=ECG+10START INTENSIVE MONITORING (record ECG readings every 1 s for the next 1 min)Violation=1; Count=0WHILE (Violation AND Countb60){ IF ECG[E]b=ECG−10 OR ECG[E]N=ECG+10Count ++; ELSE Violation=0; }IF (Violation)SEND ALARM MESSAGE TO THE ALARM AGENT AT THE E-Health Center

ELSE ECGAGENT RETURNS TO NORMAL STATE, SENDSMESG. TO THE ECGMONITOR TOCONTINUE NORMAL ECG MONITORING

Blood Sugar AgentThe Blood Sugar Agent receives blood sugar levels (BSL) via Bluetooth, checks thereceived levels for any violation. If a violation is detected then two other consecutivereadings are requested. The other readings are analyzed and if all of them meet theviolation criteria as defined by the alert conditions defined in the ontology then analarm is triggered.

IF BSL(b)b=BSL−10 OR IF BSL(b)N=BSL+10START INTENSIVE MONITORING (report two consecutive BSL readings)Violation=1; Count=0WHILE (Violation AND Countb2){ IF BSL(b)b=BSL−10 OR IF BSL(b)N=BSL+10Count ++;ELSE Violation=0; }

IF (Violation)SEND ALARM MESSAGE TO THE ALARM AGENT AT THE E-Health Center

ELSE Blood Sugar Agent RETURNS TO NORMAL STATE, SENDS MESG. TO THE BloodSugar MONITOR TO CONTINUE NORMAL MONITORING

Asthma AgentThe Asthma Agent receives peak flow levels (PFL) via Bluetooth, checks the receivedlevels for any violation. If a violation is detected then two other consecutivereadings are requested. The other readings are analyzed and if all of them meet theviolation criteria as defined by the alert conditions defined in the ontology then analarm is triggered. The alarm triggered is sent to the Alarm Agent at the E-HealthCenter.

IF PFL(p)b=PFL−10 OR IF PFL(p)N=PFL+10START INTENSIVE MONITORING (report two consecutive PFL readings)Violation=1; Count=0WHILE (Violation AND Countb2){ IF PFL(p)b=PFL−10 OR IF PFL(p)N=PFL+10Count ++;ELSE Violation=0; }

IF (Violation)SEND ALARM MESSAGE TO THE ALRM AGENT AT THE E-Health Center

ELSE Blood Sugar Agent RETURNS TO NORMAL STATE, SENDS MESG. TO THE AsthmaMONITOR TO CONTINUE NORMAL MONITORING

Medication AgentReads the signal from the Medication Sensor and sends reminders to patients to takethe medication via audio and text based messages. If medication is skipped beyond acritical level (as predefined by a healthcare professional such as: 2 dosages missed) analert is routed to the healthcare provider.

MED Missed=1, COUNT=0WHILE (MED Missed AND COUNTb4) {Send Reminders via Audio and Text to take MEDCheck Med(m) After 30 min;If Med(m) is MissedMED Missed=1;

ELSE MED Missed=0COUNT++; }.IF (MED Missed)SEND ALARM MESSAGE TO THE ALRM AGENT AT THE E-Health Center

ELSE RESUME NORMALLY.

Table 4 (continued)

Intelligent agents and corresponding decision protocols

Weight AgentReads the signal from the Weight Sensor and analyzes it. If the change in weight is ≤10units over a 24 h period sends reminders to patients regarding the weight and adviseson healthy eating habits. If the change inweight isN15 units over a 24 h period a routinealert is routed to the healthcare provider informing of the event.

IF WT CHANGEb=10 UNITSAdvice on Healthy Habits and Change in WT via Audio and Text Message

ELSE IF WT CHANGEN=15 UNITSSEND ALERT TO ALARM AGENT at the E-Health Center.

Sleep AgentReads the signal from the Sleep Sensor and sends messages to patients regarding thepattern in audio and visual mode. An alert is also routed to the healthcare providerinforming them of the event

IF SLEEPLESSNESS/TOO MUCH SLEEPAdvice via Audio and Text Message

IF SLEEPLESSNESS/TOO MUCH SLEEP FOR MORE THAN 2 CONSECUTIVE DAYSSEND ALERT TO ALARM AGENT at the E-Health Center.

616 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

from the monitors and the sensors, look for violation of pre-specifiedthresholds, and transmit alerts to the healthcare centerswhen needed.The physician specifies the thresholds, which a particular Agent willrefer to before triggering an alarm. The frequency of monitoring eachvital sign, under normal conditions (example record ECG every 10 s) aswell as abnormal conditions, i.e., after detection of an anomalousreading, (example Record ECG every 1 s for the next 1 min) is alsoconfigured. The ontology that categorizes the different alerts is built ineach PDA by a specialist that describes the alerts that must be checkedfor every patient. Table 4 gives a detail description and functionality ofthe various intelligent agents that are tasked with analyzing themonitored parameters, and the protocols.

5.1.5. What Information is transmitted and how is the signal transmitted?The sensors monitoring compliance and the devices monitoring

specific vital signs transmit the data to the patient's PDA at a pre-defined time interval. For signal transmission we propose a hybridnetwork in order to increase the reliability of transmission andenhance the network coverage. The transmission of data from thesensors and the monitoring devices to the patient's PDA is done viashort range communication medium such as Bluetooth. The transmis-sion of signals between the patients and the healthcare professionalsfor the most part is to be supported via infrastructure based wirelessnetworks such as WLAN, Cellular/PCS/GSM, and satellite basedsystems. However, the coverage of infrastructure based networksinherently suffers from temporary/permanent dead spots therebynegatively impacting the reliability of transmission [48]. Thus wepropose to use mobile ad-hoc network to complement the coverage ofexisting infrastructure based networks in areas with limited or nocoverage. Utilizing a hybrid network comprising of mobile ad hocnetwork and other infrastructure based wireless network builds alayer of fault tolerance with respect to patient–doctor communicationand thus increases the reliability of transmission. In case an alarmmessage needs to be transmitted and the patient is outside thecoverage of the infrastructure based network then an ad-hoc networkformed between the patient's PDA and other devices within the rangeof the patient's PDA can transmit the alarm via multiple hops till itreaches a healthcare professional or is picked up by an infrastructurebased network for further transmission to a healthcare professional(Fig. 10).

5.2. Usage scenario

A descriptive usage scenario is presented to demonstrate the utilityof the proposed conceptual model. Patient X is being continu-ously monitored for detection of anomalous conditions and disease

Fig. 10. Transmission from patient to healthcare professional via hybrid network.

617S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

management. One morning the patient woke up at 6 am. The sleepsensor did not observe any irregularity in patient's sleep patterns.The patient goes to the restroom and takes his weight. The weightsensor observes no critical change in patient's weight. The patient isgetting ready to go to out and has forgotten to take his morningmedication. The medication sensor senses that the medicationhasn't been taken by 8 a.m. and sends a signal to the medicationagent which reminds the patient to take the medication. The patienttakes the medication and goes to a nearby shopping mall for lunch.During the monitoring the ECG agent detects an irregularity andrequests intensive monitoring for the next 30 s. All ECG readings inthe intensive monitoring session were in the critical range. Hencean emergency alert is transmitted by the transmission agent. Thepatient is in a dead spot where there is no coverage of the wirelessnetwork. Hence an ad hoc network is formed and signal is trans-mitted from the patient's device (at themaximumpower level) to theother co-operating devices in the range of transmission. After twohops, the third hop is a cellular base stationwhich picks up the signaland routes it to the healthcare provider. The healthcare professionalreads the message, takes prompt medical action to resuscitate thepatient. In the absence of ubiquitous patient monitoring, the patientcould have lost his life and/or ended up in the hospital for anextended period of time due to complications arising from missedmedications, or anomalies going unnoticed. The key contribution ofubiquitous patient monitoring is improved healthcare delivery bytimely and reliable detection of anomalies and enhancing theefficiency of the physicians by assisting them in providing pertinentmedical attention as and when needed.

6. Conclusion and future research

Technology based ubiquitous patient monitoring has been pro-posed for promoting wellness, prevention, disease management,compliance, and reduced incidences of hospitalizations and corre-sponding expenses [10,18,33]. However, the focus so far has been onthe development of artifacts with little or no attention given torefining the process of patient monitoring and defining clear guide-lines that can be universally applied to developing effective, efficientpatient monitoring solutions. In the current research we address this

gap and make the following contributions to the existing body ofresearch in patient monitoring. First we develop a framework thatcharacterizes the basic processes that can be universally applied forubiquitous patient monitoring. Second, we describe in detail thedecision protocols, key parameters, and enabling technologies foreach process. Third, we articulate the requirements and challenges ofubiquitous patient monitoring and evaluate the myriad of technolo-gies enabling the development of ubiquitous patient monitoring. Theproposed framework and decision protocols are grounded in therequirements of patient monitoring. The requirements, challenges,and evaluation of enabling technologies are based on an in depthreview of pertinent literature in the area. Lastly, we also propose aconceptual model of a ubiquitous patient monitoring system based onthe proposed framework and clear concise guidelines. The contribu-tion of the proposed innovative strategy is effective healthcaredelivery, increased compliance with medical advice, and improvedquality of life of patients outside of the hospital.

Although, ubiquitous patient monitoring is still in its infancy thepossibilities of leveraging it as a decision support tool are vast and therealization process has merely begun. Future work will help to finetune the concept and help to bring forth the realization of a ubiquitoushealthcare environment. It is our hope that some of the current issueswill open the door for future research. Future work may address theissue of issue of acceptability of the system along with the level oftrust of doctors and patients in the ubiquitous healthcare environ-ment. A research based on testable hypotheses can potentially providesome substantive finding with respect to adoption of ubiquitoushealthcare. In the light of HIPAA, security and reliability of transac-tions made over a wireless network are important concerns. Futureresearch addressing the issue of increasing security and reliability ofmobile communication is imperative to the success of ubiquitoushealthcare. There is also research needed to address the mechanismexploring how patient monitoring systems can be utilized as adecision support tool for healthcare professionals.

Acknowledgment

The work was supported, in part, by a National Science Foundation(NSF) research grant (SCI#0439737).

618 S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

References

[1] P. Bonato, Wearable sensors/systems and their impact on biomedical engineering,IEEE Engineering Medicine and Biology Magazine 22 (3) (2003).

[2] O.J. Bott, E. Ammenwerth, B. Brigl, P. Knaup, E. Lang, R. Pilgram, B. Pfeifer, F. Ruderich, A.C.Wolff, R. Haux, C. Kulikowski, The challenges of ubiquitous computing in health care:technology, concepts and solutions, Methods of Information in Medicine 44 (3) (2005).

[3] C. Boult, R.L. Kane, J.T. Pacala, E.H.Wagner, Innovative healthcare for chronically ill olderpersons: results of anational survey, TheAmerican Journal ofManagedCare5 (9) (1999).

[4] S. Brahnam, C.F. Chuang, R.S. Sexton, F.Y. Shih, Machine assessment of neonatalfacial expressions of acute pain, Decision Support Systems 43 (4) (2007).

[5] M. Chiasson, E. Davidson, B. Kaplan, R. Kukafka, G. Kuperman, Strangers in a strangeland: can IS meet the challenges and opportunities of research in healthcare?Proceedings of the Tenth Americas Conference in Information Systems, 2004.

[6] Chronic Care Improvement, ITAA E-Health White Paper: A Product of the E-healthCommittee, 2004, p. 5.

[7] R.C. Coile, B.E. Trusko, Healthcare 2020: technology in the newmillennium, HealthManagement Technology 20 (11) (1999).

[8] J.M. Corchadoa, J. Bajo, Y. Paza, D.I. Tapiaa, Intelligent environment for monitoringAlzheimerpatients, agent technology forhealthcare,DecisionSupport Systems44 (2)(2008).

[9] M.J. Covington, W. Long, S. Srinivasan, A.K. Dev, M. Ahamad, G.D. Abowd, Securingcontext-aware applications using environment roles, In Symposium on AccessControl Model and Technology, 2001.

[10] E. Dishman, Inventingwellness systems for aging inplace, IEEE Computer 37 (5) (2004).[11] G.B. Doherty, R.N. Ross, P.E. Ross, The effectiveness of an interactive electronic lung

function monitoring system in the total management of refractory asthma, DiseaseManagement and Health Outcomes 3 (2) (1998).

[12] I.A. Gieras, The proliferation of patient-worn wireless telemetry technologies withinthe U.S. healthcare environment, Proceedings of 4th International IEEE EMBS SpecialTopic Conference on Information Technology Applications in Biomedicine, 2003.

[13] S. Goldberg, N. Wickramasinghe, 21st century healthcare — the wireless panacea,Proceedings of the 36th Hawaii International Conference on Systems Sciences, 2003.

[14] F. Gouaux, L. Simon-Chautemps, S. Adami, M. Arzi, D. Assanelli, J. Fayn, M.C. Forlini,C. Malossi, A. Martinez, J. Placide, G.L. Ziliani, P. Rubel, Smart devices for the earlydetection and interpretation of cardiological syndromes, In Proceedings of the 4thInternational IEEE EMBS Conference on Information Technology Applications inBiomedicine UK, 2003.

[15] P.J.H.Hu, C.P.Wei, T.H. Cheng, J.X. Chen, Predicting adequacyof vancomycin regimens:a learning-based classification approach to improving clinical decision making,Decision Support Systems 43 (4) (2007).

[16] K. Hung, Y. Zhang, Implementation of a WAP-based telemedicine system for patientmonitoring, IEEE transactions on Information Technology in Biomedicine 7 (2) (2003).

[17] D.J. Hunter, Disease management: has it a future? 320 (7234) (2000).[18] IOM. To err is human: building a safer health system, US Institute ofMedicine Report.

http://www.nap.edu/books/0309068371/html/.[19] R.S.H. Istepanian, A.A. Petrosian, Optimal zonal wavelet-based ECG data compression

for a mobile telecardiology system, IEEE Transactions on Information Technology inBiomedicine 4 (3) (2000).

[20] S. Jain, Energyaware communication inad-hocnetworks, Technical ReportUniversityof Washington Seattle–CSE, March 2003.

[21] E. Jovanov, A. O'Donnel, A. Morgan, B. Priddy, R. Hormigo, Prolonged telemetricmonitoring of heart rate variability using wireless intelligent sensors and a mobilegateway, In Proc. Second Joint IEEE EMBS/BMES Conference, 2002.

[22] E. Jovanov, A. O'Donnell Lords, D. Raskovic, P.G. Cox, R. Adhami, F. Andrasik, Stressmonitoring using a distributed wireless intelligent sensor system, IEEE Engineer-ing in Medicine and Biology Magazine 22 (3) (2003).

[23] E. Kafeza, D.K.W. Chiu, S.C. Cheung,M. Kafeza, Alerts inmobile healthcare applications:requirements and pilot study, IEEE Transactions on Information Technologies inBiomedicine 8 (2) (2004).

[24] A. Kara, Protecting privacy in remote-patient monitoring, IEEE Computer 34 (5) (2001).[25] S.E. Kern,D. Jaron, Healthcare technology, economics and policy: anevolvingbalance,

IEEE Engineering in Medicine and Biology Magazine 22 (1) (2003).[26] S. Khoor, K. Nieberl, K. Fugedi, E. Kail, Telemedicine ECG-telemetry with Bluetooth

technology, Proceedings of Computers in Cardiology, 2001.[27] I. Korhonen, J.E. Bardram, Guest editorial introduction to the special section on

pervasive healthcare, IEEE Transactions on Information Technology in Biomedicine8 (3) (2004).

[28] I. Korhonen, J. Parkka, M.V. Gils, Health monitoring in the home of the future:infrastructure and usage models for wearable sensors, IEEE Engineering Medicineand Biology Magazine 22 (3) (2003).

[29] J.C. Kyu, H.H. Asada, Wireless, battery-less stethoscope for wearable healthmonitoring, Proc. the IEEE 28th Annual Northeast Bioengineering Conference, 2002.

[30] L. Lin, P. J-HHu,O.R.L. Sheng,Adecision support system for lowerbackpaindiagnosis:uncertainty management and clinical evaluations, Decision Support Systems 42 (2)(2006).

[31] K. Liszka, M. Mackin, M. Lichter, D. York, D. Pillai, D. Rosenbaum, Keeping a beat onthe heart, IEEE Pervasive Computing 3 (4) (2004).

[32] A. Lymberis, A. Smart,Wearables for remote healthmonitoring, from prevention torehabilitation: current R&D, future challenges, Proc. 4th International IEEE EMBSConference on Information Technology Applications in Biomedicine, 2003.

[33] M.K. McGee, E-Health on the Horizon, 989, May 2004.[34] G.G.Mendoza, B.Q. Tran, In-homewirelessmonitoring of physiological data for heart

failure patients, In Proc. of the Second Joint IEEE EMBS/BMES 24thAnnual Conferenceand the Annual FallMeeting of theBiomedical Engineering Society Conference, 2002.

[35] D. Nussbaum, X.Wu, An architecture of scalablewirelessmonitoring system, In Proc.of the 25th Annual international Conference of the IEEE EMBS CancunMexico, 2002.

[36] L. Qin, T. Kunz, Survey onmobile ad hoc network routing protocols and cross-layerdesign, Carleton University Systems and Computer Engineering, Technical ReportSCE-04-14, August 2004.

[37] R. Rajaraman, Topology control and routing in ad hoc networks: a survey, ACMSIGACT News, 33(2) (2002).

[38] A. Sarela, I. Korhonen, L. Lötjönen, M. Sola, M. Myllymäki, IST Vivago— an intelligentsocial and remote wellness monitoring system for the elderly, In Proceedings of the4th Annual IEEE EMBS Special Topics Conf. on InformationTechnology Application inBiomedicine, 2003.

[39] M. Satyanaryanan, Pervasive computing: vision and challenges, IEEE PersonalCommunications 8 (4) (2001).

[40] S. Sesay, Z. Yang, J. He, A survey on mobile ad hoc wireless networks, InformationTechnology Journal 3 (2) (2004).

[41] S. Sneha, U. Varshney,Wireless ECGmonitoring system for pervasive healthcare, InProceedings of the Eleventh Americas Conference on Information Systems, AMCIS,Omaha NE, 2005.

[42] S. Sneha, A. Dulipovici, Strategies for working with digital medical images, InProceedings of Hawaii International Conference on Systems Sciences, 2006.

[43] S. Sneha, Patient Monitoring via Mobile Ad Hoc Networks: Maximizing Reliabilitywhile Minimizing Power Usage and Delays, Doctoral Dissertation, Georgia StateUniversity, May 2008.

[44] V. Stanford, Using pervasive computing to deliver elder care, IEEE Pervasive ComputingMagazine 1 (1) (2002).

[45] T. Suzuki, M. Doi, LifeMinder: an evidence-based wearable healthcare assistant,Proc. ACM CHI Conference, 2001.

[46] A. Tablado, A. Illarramendi, J. Bermudez, A. Goni, Intelligent monitoring of elderlypeople, Proc. 4th Annual IEEE Conf. On Information Technology Applications inBiomedicine UK, 2003.

[47] N.H. Vaidya, Tutorial on Mobile Ad Hoc Networks: Routing, MAC and TransportIssues, INFOCOM, 2006.

[48] U. Varshney, Pervasive healthcare: applications, challenges and wireless solutions,Communications of the Association of Information Systems, 2005.

[49] U. Varshney, A framework for supporting emergency messages in wireless patientmonitoring, Decision Support Systems, April 2008.

[50] Website:AwareHomeProject:http://www.awarehome.gatech.edu/projects/index.html.[51] Website: Biotronik, http://www.biotronik.com/.[52] Website: Cardionet, http://www.cardionet.com/.[53] Website DAML: http://www.daml.org (2005).[54] Website Description Logic Homepage: http://dl.kr.org (2003).[55] Website: Gator Tech Smart House: http://www.icta.ufl.edu/gt.htm.[56] Website: LifeSync — www.wirelessecg.com, 2005.[57] Website: Patient Monitoring by Welch Allyn (http://www.monitoring.welchallyn.

com/products/wireless/).[58] Website for Smart Shirt: http://www.smartshirt.gatech.edu.[59] Website: TV-based Monitoring by Philips (http://www.medical.philips.com/main/

news/content/file_630.html.[60] Website: US Administration on Aging Report on Demographic Changes, http://

www.aoa.dhhs.gov/aoa/stats/aging21/demography.html.[61] Website: US Patient Monitoring Industry Overlook. http://www.frost.com/prod/

servlet/report-homepage.pag?repid=A369-01-00-00-00.[62] M. Weiser, The Computer for the Twenty-First Century, Scientific American, 265 (3)

(1991).[63] World Health Organization, Death by Cause, Sex and Mortality Stratum in WHO

Regions, 1999.

Sweta Sneha is an Assistant Professor at Kennesaw StateUniversity in the Department of Computer Science andInformation Systems. She has a Bachelor of Science inComputer Science from University of Maryland, College Parkand a PhD in Computer Information Systems from GeorgiaState University. Sweta's research interests center around awide array of technical and behavioral challenges related tothe emergingfield of “E-Health,”which lies at the intersectionof telecommunication, information technology, healthcaresector, and business. Her goal is to research, analyze, andrecommend technical and/or behavioral solutions to meetthe challenges associated with the spiraling healthcareexpenses, the aging population, and the need to integrate/e anddeliveryof healthcare.Within the e-health umbrella she

use technology in thepractic

has conducted and published research in: (a) wireless network and enhanced decisionsupport systems for innovative e-health services seeking to economize human/financialhealthcare resources, (b) adoption, usage, and integration of emerging e-health services inthe practice and delivery of healthcare by the healthcare professionals such as: ElectronicMedical Records (EMR) and the corresponding performance improvement in the quality ofpatient care, and (c) organizational impact and process change associated with theintegration and usage of e-health services by the healthcare sector. She has publishedseveral research papers in premier IS conferences and journals including AMCIS, HICSS,IEEE Broadmed, and IEEE Communications. She has also worked as an InformationTechnology (IT) Consultant within the Management Consulting Practice of Pricewater-houseCoopers, one of the leading IT and management consulting firms in the world.

merc

619S. Sneha, U. Varshney / Decision Support Systems 46 (2009) 606–619

Upkar Varshney is on the faculty of CIS at Georgia StateUniversity. His current interests include wireless networks,

e. According to scholar.

pervasive healthcare, ubiquitous computing and mobilecommerce. He is the co-founder (with Prof. Imrich Chlamtac)of International Pervasive Health Conference and also co-chaired the conference in 2006. Upkar is also the program co-chair for Americas Conference on Information Systems(AMCIS-2009) in San Francisco.He has authored over 120 papers including about 60 injournals, including 30 in IEEE and ACM publications. He is the

author of several heavily downloaded and cited papers inwireless networks, pervasive healthcare, and mobile com-google, the total number of journal and conference citations for

his papers exceeds 1700.Upkar has presented several very well received tutorials and workshops (and even a fewkeynotes) at wireless, computing, and information systems conferences. He has alsoreceived grants from several funding agencies including the National Science Foundation.His teaching awards include Myron T. Greene Outstanding Teaching Award (2004), RCBCollege Distinguished Teaching Award (2002), and, MyronT. Greene Outstanding TeachingAward (2000).He is serving or has served as an editor/guest editor for IEEE Transactions on IT inBiomedicine, ACM/Kluwer Mobile Networks (MONET), IEEE Computer, Decision SupportSystems (DSS), Communications of the AIS (CAIS), Int. J. on Network Management (IJNM),Int. Journal on Mobile Communications (IJMC), Int. Journal of Wireless and MobileComputing (IJWMC), and Handbook of Research on Mobile Business.