Emergency care management with location-aware services

Emergency

Care

Management

with Location-Aware

Services

Shih-wei Lee* ,

Shao-you

Cheng*,

Jane

Yung-jen Hsu*,

Polly Huangt, Chuang-wen

You*

* Department ofComputer Science andInformation Engineering

National Taiwan

University,

Taiwan

{r94094, r93070, yjhsu, f91023}@csie.ntu.edu.tw t Department of ElectricalEngineering, National Taiwan

University,

Taiwan

phuang@cc.ee.ntu.edu.tw

Abstract- As indoor localization technologies become need is themost critical.

Therefore,

wetry to

implement

more affordable and sophisticated. An emerging market and experiment a set of location-aware services to as-has yet to be realized commercially in the indoor envi- sist in the management of emergency department (ED). ronment. We think that the first adopter will arise from This paper introduces an integrated project that brings

the healthcare domainwhere the need is the mostcritical. m

Thus,Weaimtoimplementasetof location-awareservices

. I.

o

toassist in the management of the emergencydepartment. atNTU

Hospital,

the

engineers

ofIBM

Taiwan,

and the

We apply NTU Taroko, an active RFID module for to researchers in NTU iSpace Lab. The three parties support

real time location tracking on patients, hospital assets, respectively the application, service, and infrastructure

and medical staffs. In addition,weintegratecontext-aware aspects of the ERmanagement system.

system proactively to infer event notifications for remind- The statistical data shows that NTU Hospital treats ing physicians and nurses. With the proposed system, we about 250 emergency patients a day. This amount is can shorten the process of emergencyvisit effectively and much more than many other countries. However, only improve the quality ofemergency care. one-fourth require admission or stay for observation.

With

triage,

the critical

patients

would be served

firstly,

I.

INTRODUCTION but most of the non-emergent patients feel dissatisfied Outdoor localization systems such asGlobal Position- after a long-term wait. Halfof them want to be treated ing System (GPS) have enabled a number of commer- in ten minutes, only

11%

could accept 30 minutes

cial location-aware services, for

example,

the vehicular wait [4]. Because of patients are many andtheir expected

navigation systems that are routinely used by many waiting time is short, how to speed up the process of

drivers today. An

emerging

market that has yet to be emergency visit becomes primary subject to improve

realized commercially is at the location-aware services the patients' satisfaction. We learn from doctor's de-for indoor environment. As indoor localization technolo- scription that nursing staff often spend much time in

gies becomemoreaffordable and sophisticated, location- findingpatients, hospital assets (e.g. sickbeds,pumps, or awareservices for indoor environments areofgreatcom- wheelchairs). If we can track their position in the ED,

mercialization potential. To some

degree,

commercial then we can spend less time to find particular patients applications such as

building security

and asset

tracking

and have better utilization of medical resources. This fall also into the location-aware service category. also means we shorten theprocessofemergencyvisit. In

While many of the indoor location-aware

applications addition, patient

and staff

tracking

enable the chances of have beenprototyped andprovenuseful. There aremany

providing

various of location-aware servicesto

improve

indoor positioning solutions available on the

market,

the qualities ofemergency care. This is the main idea such like Ekahau [1], Sonitor [2], or Versus [3], etc. for us to construct a location-aware ED management

These companies apply WiFi,

ultrasound/RFID,

or in- system.

frared/RFID technologies in real time

locating

systems 1. SCENARIO

and have

successfully

deployed

these systems in the

Triag Patientsareservedaccordingtoseverity

-~~~~~~~~~~~andacuity,notbyorder. EA R TSa

...L evel...C ondition... elC ndon..Waiting.. ...T.

...Peio d.... try... ...Several-ill....or..injured...whoPehas ...Level..2.stable...vital..signs..but..need..10..mins Physiian.ork.p.urgnt.mdica.car_{or...injured....patients....who...need.~~ ~ Il} Level 4

clinic-Treatment~~~~~~~~~~ptin

Thepat entwith inunresledo wohsymtm a

Discharge to stay in the observation unit...

Fig... 1..The..process..of.emergency...visit.at..NTU.Hospital..Fig...2..Deployment

Lvof

thealocation

sgsystem

ind

themED

"Bob...was.involved...in..a.car..accident...and..hurt..his..legs...of.location...system...in.the..ED..at NTU..Hospital....In..our

He..was..sent..to..N...Hospital...by..the..ambulance....Upon..current...prototype,...the..location...system..supports...location. urgetptiet ad tgge wih a aciveRFI bage.

Thetrakedasstsgnclde

ickedsandwhelchirs

the on-calldoctor

who~~~~~~~

wase

takloringjcredo tepatients

localization.n

The M

physicm

ian and nussalowaratv in the observation unit, and

~~~Sinfoereamedt

hrtat

aut

patient RDAbadge for thkeirP siftor vii.Eahptendol

justregistered and

need~~~~~~~~edvae.

Wihot

>n

dela

Jane

be.. assigned.... anatv..D ag fe amfigh

examines....

Bob,..and..orders...for..an..X-Ray..to..be..taken....badge..has..ID..number...that..is..associated....to.the..patient.ni

In..the.. meantime,...Polly..retrieves...the..nearest..wheelchair...The..emergency...physicians...and..nurses..can..also..use..their

available...with..the..help..of.the..location...system...After..the..badges..as..identification...to..log.into..the..emergency...care while...waitingHHHHHfor.th.X-Ra.outcme..Twnty.mnutes succesful.lgin.he.meical.taffs.an.acess.lcatio

late,te mnagmentsysem.ot..e.Janeas.oonas.nfomatin.acoringto.heirrols.o.shfts

Bob...s.XRay.ecoes.aailble.Jan.dianoss.tht.Te.R.D.bdgesbascall.ar.R..eceverstha

Bob...HM4

hasaisurefrctrebasd.n.heX-Ry.eslt.peioicllylite.fr.te.igal.frm.heRI.trnseier

Bob... is.mmditey. octe.ad.rete.wthstbiizng cale.bacn.Afercoletig.hesinas,th.bdg

casts... on ...his...legs...Aftr.pyin.th.meica.exenss.ad.sndsthedat.tan te...eevr,cle.rly h thehospital."

infrastructuresileexaintio toluesdra

the...F 3loctin ngnewhc detemie

(taggingand untagging

~~~~~~~The

patient)

oftunrsinglted

sysptems

ias

ho.t.omut bag'spsiin.sgie.i.ecin.

minimalcompared

wi~~~~~~th

stheitotlhmonfsrato time. for the B.o

nurse to look froresvrous

thirgngyvsitandfo

NThe

docptorlt

Contex

isDpoyeto

inomtonctatinste

the can

thel

copue disove test results. The

effiacideny

ofd

emergency

caegs

flcto

systemsocarceize

the

uEr'

sitNUation,suchtas. theou

andiflow can

bhe

grepiatly

impovedwby

elasimindatls rcing

suhnsr'

hysiclpoital,

andts memoional inaformad

ation.the

unnecessanryedselay

wtthhe

location-awarie

servicaesJn,

mhyain cotexasg

fegrfred

tihatveRI odteusers

loc-ation

the n-cll octo wh wa

taing

areof therpatent

lactiviyaietity,

andtie

[5].can

an

syrstems lontextawartie

inIhIbsraio.nt,adi

EMERENCCAErMNAGMEN SYTE

ifipueaonettopoidieevnenfraiotn

Toenablethe improed

scenaifo,h

rooe RsericeD t

thdeso

useir.

shnfour

ED

simanageen

systiem,weul

managhemeantisysem

canl betviewedh cnaeptuahellyhasia

dheveo

eeaemergency

chscasarenuservcsbanasedone

thei

interatin otwosystms,the ocaion yste an the conext espiectiailcation olointext toe improvencpatien

We acquire location context by

utilizing

the location system, however, we are not limited to use

single

type

of context. We have considered other useful contexts, X _g

e.g.

time,

patient's

treatment status, staff's

activities,

test

progress,

etc. In Section Il-A, we introduce three

common

problems

in the ED and

respectively

develop

their context-aware solutions.

.1...

A. Emergency Care Services

Fig. 3. Main components of theproposed emergencycare manage-First, "'uninformed discharges"' usually troubles the ment system

medical staffs, i.e. apatient who leaves ED

prematurely

without

completing

the

regular

ER

procedures.

The

main reason is due to

long-term

wait [6]. To manage treatment instructions, software events from

HIS1

and

such accidental

situations,

we

develop

a location-aware

PACS2,

and temporal information.

notification service to support detection of uninformed

discharges. Our system

continuously

monitor

patients'

IV. SYSTEM COMPONENTS

location and their treatment status, once such situation The

proposed

emergency care

management system

occurs, the notification will

immediately

inform the is

composed by

five

components,

including

hardware

medical staffs. _{and software. As}

_Fig.

₃

_shows,

_{these components are:}

Second, the emergency

patients

expect to be treated

(1)

active RFID module, (2) NTU Taroko Localization

as soon as

possible,

but most of them are not emergent Service, (3)

IBM

Enterprise Location Aware Service cases. To control their

staying

time

clearly,

the system Technology, (4) Context-aware Reasoning Agent, and

continuously monitors

patients'

room-level location in (5) Web-based User Interface. We further describe each

the ED and also their treatment status to estimate their component's technical details as follows.

waiting time. Ifthere are any

patient

who has

long-term

staying without

taking

anymedical

service,

their location A. Active RFID Module

will be

displayed

on the map to remind the emergency The

passive

RFID

technology

allows identification

nurses. Moreover, as the

study

suggests

[7],

the more of

objects

and persons in short

sensing

range. Limited responses andconcernsthe medical staffs

give,

the more

by

the radio range, the

passive

RFID

technology

can

satisfaction the

patients

feel. Based on this

idea,

we de- not

support

indoor localization in our emergency care

velop atime-based notification that

proactively

informs

applications.

Therefore,

wechoose active RFIDtoenable

the medical staffs when there are

patients'

waiting

time

exceeding30minutes.

~~~~both

identification and

localization

forthe

location-aware

exceeding 30 minutes. services.

Third,

additional clinical examination is

quite

often Taroko is a ultra low power, low cost, wireless em-in the emergency

visit,

however,

such test often results bedded device. The module consists of a

TI

MSP430

in some delays. For

example,

a

patient

who needs to

microcontroller,

an IEEE

802.15.4/Zigbee

radio,

and

take X-Ray, but can not find where to take it.

Maybe

1024kB size external flash memory. The

MSP430

micro-the

patient

completes

the

examination,

but he does not controller operates at 8MHz and has

l0kB

RAM. The

know where to go next, etc. There are many

possible

Chipcon CC2420 Radio Transceiveroperated at 2.4GHz

situations, however,

waiting

for final results is a com- with

250kbps

data rate and is

compliant

with IEEE mon delay. Therefore, we

develop

an event notification

802.15.4/Zigbee.

The flash memory allows

developers

service that

proactively

informs the

physician

or nurses to store data such as the identification or radio

signal

when

particular

patient's

X-Ray result is available. The

strength

for location

computation.

Data collection and

notification has two ways, one is

flashing

badge's

LED communication to PC are via standard USB interface.

light, and the other is

sending

text messagesto the web- Taroko also supports

integrating

with additional sensors based

notification

board.

RSSIinfoKNesCtoain_o_i g

RSSIFirgerprint Refposition Fingedpmintgcon

RSSIFingerprnot2 Refposition2Dtbs

R.SSIFiogerpoiot3 Refpositioni3

V ~~~~~Locations

Positio ring__ Paee ittem Dataobase signal ---o Recoganitino

feceivedbv Alg6e savetimaotmpiede

fRemobile ol cationin b

nodes deitabas s

Fig. 4. RF-based location systemwith fingerprinting approach Taroko is a

programmable

module.

Developers

can

code in NesC programming language and run the

pro-

Fig. 5. IndoordeploymentofTaroko-based location

system;

beacons

gram upon TinyOS [8], a small and open-source soft- and

relays

are attached to ceilings in passages, anda badge is worn ware

operating system designed

for sensor networks. bya

person

Alternatively,

the

developer

can choose to

program

in

C and

compile

the code

by

commercial

compilers (e.g.

JAR embedded

workbench)

orGNU MSPGCC

compiler

as

badge.

Fig.

5

depicts

the

environment

setting.

The

for the MSP43 0 microcontroller. The

compiled

machine

badge regularly

collects beacon messages and forwards

code can be uploaded to Taroko via USB

port.

the average RSSI values to the location server to

com-B.rNTUaTarokodiocalizationoService

pute its location. In our preliminary experiments, the

B.hNTU

Taroko

Localrprizatswion Sesorvced

inth Fngr average accuracy is 1 to 2 meters. To compare

Taroko-The localization service

[9]

takes the

fingerprinting

based location

system

with traditional WiFi-based

(using

approach

and works in two

phases.

The first

phase

is 802.11

standard)

location

system.

The former has some

called

offline

training

phase

in which a human

operator

advantages, including inexpensive

RFID module

(about

performs

a site survey

by measuring

the received

signal

60 USD per

Taroko),

small

(about

the size ofa

cigarette-strength

indicator

(RSSI)

from different beacon nodes

case), low-power (powered by

two AAA

batteries),

and

(BNs)

at some fixed

sampling points

inthe environment,

easily programmable (programming

and

communicating

Given these RSSI measurements, one can compose a via USB

port).

radio fingerprint of BNs at different sampling points.

C.

c erprise Location Service

These radio

fingerprints

will be stored in the

Finger-

C.pIBMent

ation Aae Service Technlg

printing

Database. The second phase is known as online IBM

Enterprise

Location Aware Service

(LAS) [10]

is estimation

phase.

Each mobile node is carried by a a software middleware that leverages IBM Watson

Re-personoraffachedto an

object,

it

gathers

the RSSI values search Center's PLACES

(Points-of-interest,

Locations,

from BNs and sends the databack

through

beacon nodes and Asset

Catalog

for Enterprise

Services)

Technology.

orother

relay

nodes toa backend location server.

Finally,

It provides a set of Java APIs to

compatibly

imple-the location ofmobile node is determined

by matching

ment certain core services of

OpenGIS

Location Service the measuredRSSIvaluestothe stored radio

fingerprints. (OpenLS) implementation specification [11],

such as

The

matching

process

is

computed

by

applying

a

paffern

Presentation Service

(map

portrayal),

Location

Utility

recognition algorithm.

The location server

continuously

Service

(a

reverse

geocoder

to transform a

geometry

computes

the location ofmobile nodes. Fig. 4 shows the position to

particular place).

IBM Enterprise LAS aims

main elements in the RF-based location

system.

to

help developers

build their location

tracking

web

We have

deployed

a number of Tarokos as BNs in

applications effortlessly

and consolidate different po-both NTU CSIE

building

and IBM China Software De-

sitioning technologies (e.g.

RFID,

Zigbee,

or

UWB)

manage various spatial entities belonging to different

layers in a single map. The preferred map is in scalable

vector

graphics (SVG)

format. SVG is a XML

markup

.'~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.

..H_...

._...

language

for

describing

two-dimensional vector

graph-ics both static and animated.

By

using PlacesAdmin,

we

import

SVG mapinto the database andcreateCoordinate

Reference

System

(CRS)

for the map.

Then,

we start

to create various spatial entities to

represent

real world

objects

in different layers, like rooms, hospital assets,

and

people,

etc. -A

We develop a dynamic web site with ED's map _______

as our main user interface. In addition, we use

Ajax

technique to enhance our web application. Because of the requirement for frequently updating the position of

mobile nodes, the Ajax-enhancedwebpage allows users Fig. 6. Web-based user interface for the location-aware emergency

to select multiple pre-defined objects on the single map care

and render the map without heavy page-reloading. It is

especially useful when performing user's query, since software components to support different functionalities the query results are expected to display continuously in our emergency care management system. We apply snd

smoothly

on the map.

By

utilizing

SVG and

Ajax

event-driven architecture design based on Java Message

techniques,

our web

application

presents

objects'

loca-

..

tion in a more

dynamic

manner and

greatly

avoids the c

en(JMS

facilita

pealitycbetweentse

reloading

delay

resulted from

re-generating

the whole

comp onent

nsupports

publish/describe

mechanism

map.

~~~~~~~~~~~~for

the

reasoning

agent to

subscribe

the updated context

and relevant events from HIS or PACS. These

envi-D. Context-aware Reasoning Agent ronmental events enrich the

agent's

KB for performing

The context-aware system should automatically pro- context-aware

reasoning.

vide relevant services at

right

time and in

right place.

So

far,

we

implement

the inferred actions

by

publish-To infer

appropriated

actions in the

right situation,

we

ing

messages to any interested web-based clients in the develop a software agent that makes context-aware rea- ED over JMS. For

example,

if the

physician

wants to

soning.

The

agent

performs

reactive rule-based

reasoning

be notified after his cared

patients

takes

X-Ray.

After

by utilizing

the Context-AwareRule

Engine (CARE) [12]

he subscribes to the notification under this

particular

as its knowledge baseKBtodedu t R situation defined in

rules,

the

reasoning

agent

starts to

as

itS

knowledge base

(KB)

to

dleduce

actions.

Rule-based

programming gives developers

the

flexibility

and monitor location context when the

patient

leaves for

maintainability

to manage the

system's

behaviors

by examination,

and

finally

infers the notification while the

loading

different rule sets and facts. In

general,

we build

patient

finishes the examination and exits the

X-Ray

area.

CARE as the context-sensitive version of Jess [13], a E. Web-based User

Interface

general purpose rule engine purely developed by Java

language. Because of context rapidly changes, CARE We design a web-based user interface to display

asserts context as dynamic facts and maintains them by location information and to provide a query interface

keeping

context updates from the sensors (e.g. RFID). forusersto search forperson and medical resource. The When context information changes, it will be coded main searching page can be divided into five parts, as

and sent to the reasoning agent to update its KB. For Fig. 6 depicts, each part is labeled in capital. Next, we

example, if a patient moves from hallway to the adult describe each part and its main functionalitiesasfollows.

medicine area in the ED, such room-entrance event A is a mapthat shows whole emergency care areas in

will be detected by the location system. The agent the ED. After user selects some items in part B, C,

also displays important signs, including restrooms, Lab (CSDL) in Taiwan for their technical support of

public telephones, convenient stores, etc. IBM Enterprise LAS and PlacesAdmin. We would like

C is a specialized search bar for looking for specific to thank other members ofiSpace Laboratory, especially medical staff in the ED. In the current prototype, Prof. Polly Huang, Prof. Hao-hua Chu, and Chuang-wen we supports searching for physician, nurse, and You for their discussion and support of NTU Taroko patient. Ifthe person's role has been selected, the location system. Special thanks should go to the

pro-second combo box will retrieve their names from fessionals of the Department ofEmergencyMedicine of thehospital database and list thesenamesfor further NTU Hospital, Hui-Min Chen and Ken-Shiao Lin for

selection. their valuable suggestions.

D is an event notification board that automatically

shows the unread notification messages ofpresent REFERENCES

login user. [1] Ekahau. [Online]. Available: http://www.ekahau.com

E is a context-aware

panel

to search for user related [2] Sonitor. [Online]. Available:

http://www.sonitor.com

[3] Versus. [Online]. Available: http://www.versustech.com

information. For example, if a physician logins, it [4] M. H. Chen and Y C. Huang, "Public awareness of triage allows the physician to select his cared patients, andwaitingtime atemergency department,"JournalofTaiwan

collaborated nurses, and available assets. Emergency Medicine, vol. 5, no. 3, pp. 128-131, September

2003.

V. CONCLUSION AND FUTURE WORK [5] A. K. Dey and G. D. Abowd, "Towards a better understanding ofcontextandcontext-awareness,"GeorgiaInstitute of Technol-In this paper, we introduce the emergency care man- ogy, College ofComputing, Tech. Rep. GIT-GVU-99-22, June

agement system with location-aware servicesto improve 1999.

patientcare

andemegencyvsitflo.Bas[6]

S. J.Liaw,P. M. Hu,and H. C.Liao,"Patients who leave

emer-patient

care and emergency

visit

flow. Based on the gency departments prematurely," Journal of Taiwan Emergency

RF-based indoor location system, we support patient Medicine, vol. 4, no. 2, pp. 40-50, June 2002.

tracking, staff tracking, andassettracking. The delaycan [7] M. L. Liou, "Perceived waiting times, actual waiting times be reduced by proactive context-aware notifications. We and their relations to satisfaction ofpatients of the emergency

department in a medical center," Master's thesis, Graduate

develop an agent-based context-aware system to infer Institute of Health Care Organization Administration, NTU,

appropriate notifications and apply event-driven design Taiwan, 2001.

to better integrate these multi-functional components. [8] TinyOS. [Online]. Available: http://www.tinyos.net

[9] C. W. You, Y C. Chen,H. H. Chu,P.Huang, J. R. Chiang,and

Up until now, our system has been successfully de- S. Y Lau, "Sensor-enhanced mobility prediction for energy-ployed and experimented in NTU CSIE building and efficient localization," To appeared in IEEE SECON 2006,

IBM CSDL to fully test its functionality. To avoid Reston, VA, USA,

September

25-28 2006.

bothering the emergency

patients,

we are working on

[10]

S.

C. Shen and Y C. Chang, "Best

prac-the deal ihamnsesofDprmn

fEegtices

for location-aware services," Jun 2006. [On-the details with administers of Department of Emergency line]. Available: http://www- 128.ibm.com/developerworks/edu/

Medicine ofNTUHospitaltofindout anappropriate area wi-dw-wi-best-i.html?S-TACT=105AGX05&S-CMP=HP for real deployment and evaluations.

[I

1] The OpenGIS Location Service (OpenLS). [Online].Available:

The current prototype system deals with

simple

emer-

[12]http://www.opengeospatial.org/standards/olscore

_{[12] W. R. Jih, J. Y J. Hsu, C. L. Wu, C. F. Liao, and S. Y Cheng,} gency care scenarios to demonstrate the benefits of "A multi-agent service framework for context-aware elder

care,"

the location-aware services. Our future work includes inAAMAS'2006WorkshoponService-OrientedComputingand exploring other complex scenarios for real problems in Agent-Based Engineering (SOCABE'2006), May 8-12 2006,pp.

the ED and ealuating ho much improvment we may 61-75.

the ED and evaluating how much improvement we may [13]

E. F. Hill,

"Jess, the rule engine for the java platform,"

achieve with the emergency care services. Thought we Sandia National Laboratories. [Online]. Available: http:

incorporate a location-aware middleware into the sys-

//herzberg.ca.sandia.gov/jess

tem, we still need more experiences to fully understand

its advantages and supports of building location-aware applications.

ACKNOWLEDGEMENTS

This research was supported by IBM Taiwan and by a grant from the Advanced eCommerce Institute of the Institute for Information Industry (III) in Taiwan. We would like to thank IBM China Software Development