Australasian Emergency Nursing Journal (2014) 17, 161—166

Available online at www.sciencedirect.com

ScienceDirect

journa l h om epage: www.elsev ier .com/ l ocate /aenj

RESEARCH PAPER

The emergency department prediction ofdisposition (EPOD) study

Milan R. Vaghasiya, RN, MPH, MHM, B Nursing ∗Margaret Murphy, CNC, GradDip (Mental Health,ICU Change Management), MHScEdDaniel O’Flynn, RN, MCL (clinical leadership), PG EN, B NursingAmith Shetty, MBBS, FACEM

Westmead Hospital Emergency Department, Corner of Darcy and Hawkesbury Roads, Westmead, Sydney,New South Wales 2145, Australia

Received 21 October 2013; received in revised form 7 July 2014; accepted 8 July 2014

KEYWORDSEmergency medicalservices;Triage;Organisationalefficiency;Patient-centred care;Hospitalrestructuring;ED planning

SummaryBackground: Emergency departments (ED) continue to evolve models of care and streaming asinterventions to tackle the effects of access block and overcrowding. Tertiary ED may be able todesign patient-flow based on predicted dispositions in the department. Segregating discharge-stream patients may help develop patient-flows within the department, which is less affectedby availability of beds in a hospital. We aim to determine if triage nurses and ED doctors canpredict disposition outcomes early in the patient journey and thus lead to successful streamingof patients in the ED.Methods: During this study, triage nurses and ED doctors anonymously predicted dispositionoutcomes for patients presenting to triage after their brief assessments. Patient dispositionat the 24-h post ED presentation was considered as the actual outcome and compared againstpredicted outcomes.

Results: Triage nurses were able to predict actual discharges of 445 patients out of 490 patientswith a positive predictive value (PPV) of 90.8% (95% CI 87.8—93.2%). ED registrars were ableto predict actual discharges of 85 patients out of 93 patients with PPV of 91.4% (95% CI 83.3—95.9%). ED consultants were able to predict actual discharges of 111 patients out of118 patients with PPV 94.1% (95% CI 87.7—97.4%). PPVs for admission among ED consultants,ED registrars and Triage nurses were 59.7%, 54.4% and 48.5% respectively.

∗ Corresponding author. Tel.: +61 433920594.E-mail addresses: milan aus@yahoo.com.au, milan.aus@gmail.com (M.R. Vaghasiya).

http://dx.doi.org/10.1016/j.aenj.2014.07.0031574-6267/Crown Copyright © 2014 Published by Elsevier Ltd on behalf of College of Emergency Nursing Australasia Ltd. All rights reserved.

162 M.R. Vaghasiya et al.

Conclusions: Triage nurses, ED consultants and ED registrars are able to predict a patient’sdischarge disposition at triage with high levels of confidence. Triage nurses, ED consultants, andED registrars can predict patients who are likely to be admitted with equal ability. This data maybe used to develop specific admission and discharge streams based on early decision-making inEDs by triage nurses, ED registrars or ED consultants.Crown Copyright © 2014 Published by Elsevier Ltd on behalf of College of Emergency NursingAustralasia Ltd. All rights reserve

I

EoTpspp

fOsoswait

l

pdbRsSohissrdacwtIer

M

S

W(1tmrt(

ptochwsf

S

We conducted this prospective observational study in totalduration of 7 weeks and 3 days from 7th August 2012 to

What is known

Access block and ED overcrowding adversely affect EDfunctioning and the patient’s length of stay (LOS) inED. Emergency departments (ED) continue to evolvemodels of care and streaming as interventions to meetthe increasing demand and to tackle the effects ofaccess block and overcrowding. Most tertiary ED nowhave streamed models of care e.g. resuscitation, acutecare, fast-track or walk-in clinics and/or short-stayunits. The streaming of patients to these models ofcare occurs during various stages of the patient journeythrough the ED.

What this paper adds?

Triage nurses, ED consultants and ED registrars are ableto accurately predict patient’s discharge disposition attriage with high levels of confidence. Triage nurses, EDconsultants, and ED registrars can also predict patientswho are likely to be admitted with equal ability. Thisdata may be used to develop specific admission and dis-charge streams based on early decision-making in EDsto facilitate patient-flows during the periods of accessblock and overcrowding.

ntroduction

mergency departments (ED) are introducing novel modelsf care to more efficiently meet the rising service demands.he unpredictable clinical nature of patient presentationsoses a significant challenges for planning in ED organi-ation. Despite the unpredictable clinical nature of theseatients, disposition trends such as admission rates to hos-ital appear to follow predictable trends.1

Access block and ED overcrowding adversely affect EDunctioning and the patient’s length of stay (LOS) in ED.11

bservation wards and short-stay units have added newtreaming options for ED patients who require a short periodf admission for clinical observation and intervention.2 EDhort-stay units have traditionally been reserved for patientsho are most likely to be discharged within 24 h. Medicaldmission and assessment units have become commonplacen many hospitals as a conduit for medical admissions from

he ED3,4.

Recent studies indicate that triage processes haveimited role for dimensions such as case-complexity and in

2sm

d.

redicting hospitalisation rates.5,6 Studies in the past haveocumented improved ED efficiency with streaming patientsased on case-complexity7,8 and disposition outcomes.9,10

ecent studies have shown positive results based on earlytreaming decisions as part of a multi-pronged approach.11

imilarly, past studies have also depicted the capabilitiesf different emergency services in predicting the need forospital admission.12—15 However, no studies to date havencluded early streaming by senior ED doctors at triage oroon after triage to predict a patient’s disposition. In thistudy, we aimed to determine whether ED consultants, EDegistrars and triage nurses can accurately predict patientisposition outcomes early in a patient’s journey in ED, i.e.dmission to inpatient hospital unit or discharge. This studyan help to find ways to deal with overcrowding and longaiting time in emergency rooms by streaming patients to

heir disposition based facility early in the patient’s journey.f accurate patient disposition outcome could be determinedarlier, then this may result in more effective use of EDesources and enhance patient safety and satisfaction.

ethods

etting

estmead Hospital is a tertiary 850-bed adult hospital650 beds available for emergency admissions) with a level

trauma service and 41 treatment spaces in ED, whichreats over 60,000 adult patients annually. Westmead ED hasultiple models of care including acute treatment areas,

esuscitation bays, a low-acuity fast-track urgent care cen-re, Senior Assessment Further Evaluation after Triage ZoneSAFE-T zone)11 and a short-stay unit.

In current practice, a triage nurse initially assessesatients presenting to Westmead ED. Patients meeting Aus-ralasian Triage Scale (ATS) 1 and 2 are immediately placedn an acute care bed when available. The majority of ATSategories 3—5 are reviewed in SAFE-T zone (functioningours 1000—2200 h) by ED consultants and ED registrarsho initiate investigations and management plans prior to

treaming patients to the appropriate model of care i.e.ast-track, discharge, acute or short-stay unit.11

tudy design

7th September 2012 between 1000 h and 2200 h. During thistudy triage nurses, ED consultants and ED registrars anony-ously recorded their patient’s predicted disposition on the

The EPOD study 163

Table 1 Predefined prediction outcome criteria for E-POD study.

Prediction Success criteria Failure criteria

Admission Patient admitted to in-patient unit in hospital Patient discharged from ED or from hospital within 24 hDischarge Patient discharged home or admitted to ESSU

during the visitPatient admitted to in-patient unit in hospital ortransferred to another hospitalPatient LOS in ED or ESSU of more than 24 hPatient representing within the 48 h period after beingdischarged

Short-stay unit Patient discharged home from ED when ESSUring

Patient LOS in ED or ESSU of more than 24 h

S

UIt(d

S

WDpfi

R

AittfAd

ctn

poo(s

p1Aap

p

space not available or admitted to ESSU duthe visit

provided data forms. Each data form contained the follow-ing details: (1) patient label, (2) date and time of predictionmade, (3) triage category of the patient, and (4) predictions(admission, discharge, short stay). Staff was only allowed tomake one prediction for each patient meeting triage cate-gories 3—5. Staff recorded their prediction on a data formand dropped them in a secure box. We blinded the predic-tions by triage nurses from the predictions by ED consultantsas well as ED registrars and vice versa.

Triage nurses made prediction of disposition outcomesoon after they triaged patients and recorded their predic-tions in the data forms. Followed by triage, ED registrars orED consultants made a quick assessment of 10 min on thesepatients in SAFE-T area and filled separate data forms fortheir predictions. In this assessment, ED registrars and EDconsultants gathered patients’ brief history and attendedphysical examinations.

We excluded the patients with ATS categories 1 and 2due to their acuity of presentation and need to bypass theSAFE-T zone. We also excluded patients who were referredfrom out-of-hospital with a clear plan of management i.e.planned admissions, imaging or pathology results with aclear diagnosis needing admission. The predictions werecoded under three categories namely admission, dischargeor emergency short stay (ED stay up to 24 h) to reduce inter-vention bias during the study. This was done to reflect thecurrent practices of streaming within the department. Wecrosschecked the discharged patients for a repeat presen-tation within 72 h of their initial presentation. We collecteddata forms from the secure box and collated them at theend of each week. We reviewed actual outcome data a weekafter the end of the study period. There were no failed dis-charges or representations of the patients within 72 h. Wedid not make direct contacts with patients, general practi-tioners or other hospitals during or after the study period tocheck for representations within 48 h.

Actual outcome data were refined into two groups namelyadmitted and discharged patients (discharge and short-stayunit). Table 1 outlines the predefined outcome endpoints ofthe study.

We obtained the ethical approval from Westmead Hos-pital Human Research Ethics Committee for this studywith the approval number HREC2012/6/6.1(3541)AU RED

LNR/12/WMEAD200. Since this was an observational studyand as we did not make any change to patient’s clinical man-agement, patient’s consent for the study was waived by theethics committee.

3Aat

Patient representing within the 48 h period after beingdischarged

ample size

sing nQuery advisor version 7.0 (Statistical Solutions, Corkreland), a sample of 400 patients was considered sufficiento detect a 90% agreement with ±5% confidence intervalCI) between predicted disposition and actual disposition foroctors and nurses.

tatistical methods

e used StatsDirect statistical software version 2.7.8 (Stats-irect Ltd., Cheshire, UK) to construct 2 × 2 tables ofredictions and outcomes to calculate sensitivity, speci-city, predictive values and likelihood ratios.

esults

s shown in Fig. 1, a total of 1255 patients were coded dur-ng this study. We excluded 27 patients on the basis thathey were triage category 2 leaving 1228 patients. ED consul-ants, ED registrars and triage nurses predicted the outcomeor 237 patients, 196 patients and 795 patients respectively.

total of 168 patients were assessed by both nurses andoctors.

Table 2 outlines the predicted outcome vs the actual out-ome of the predictions made in this study. Table 3 outlineshe statistical analysis of the predictions made by triageurses, ED consultants and ED registrars during this study.

ED consultants predicted outcome for a total of 237atients. Out of those, there were 119 (50.2%) predictionsf admission and 118 (49.8%) predictions of discharge. Totalf 71 (59.7%) patients had actual admissions, whereas 11194.1%) patients had actual discharges at 24 h from the pre-entation to ED.

ED registrars predicted outcome for a total of 196atients. Out of 196 predictions, admission predictions were03 (52.5%) and discharge predictions were 93 (47.5%).mong these predictions, there were 56 (54.4%) actualdmissions and 85 (91.4%) actual discharges at 24 h from theresentation to ED.

Triage nurses predicted outcome for a total of 795atients. Out of 795 predictions, admission predictions were

05 (38.4%) and discharge predictions were 490 (61.6%).mong these predictions, there were 148 (48.5%) actualdmissions and 445 (90.8%) actual discharges at 24 h fromhe presentation to ED.

164 M.R. Vaghasiya et al.

Figure 1 Consort flow diagram of patients in the E-POD study.

Table 2 Cross-tabulation of predicted outcome and actual outcome for consultants, registrars and triage nurses.

Actual Outcome Total

Admission Discharge

Predicted Outcome

ED consultants Admission 71 (59.7%) 48 (40.3%) 119 (100%)Discharge 7 (5.9%) 111 (94.1%) 118 (100%)Total 78 (32.9%) 159 (67.1%) 237 (100%)

ED registrars Admission 56 (54.4%) 47 (45.6%) 103 (100%)Discharge 8 (8.6%) 85 (91.4%) 93 (100%)Total 64 (32.6%) 132 (67.4%) 196 (100%)

Triage nurses Admission 148 (48.5%) 157 (51.5%) 305 (100%)Discharge 45 (9.2%) 445 (90.8%) 490 (100%)

al

trp55C

t4ftc8

ftdsd

D

Tot

As shown in Tables 2 and 3, the proportion of true posi-ives and false positives for admission was 59.7% and 40.3%espectively among ED consultants. The proportion of trueositives and false positives for discharge was 94.1% and.9% respectively. PPV for admission was 59.7% (95% CI0.3—68.4) compared to PPV for discharge of 94.1% (95%I 87.7—97.4%).

Among ED registrars, the proportion of true posi-ives and false positives for admission was 54.4% and5.6% respectively. The proportion of true positives and

alse positives for discharge was 91.4% and 8.6% respec-ively. PPV for admission was 54.4% (95% CI 44.3—64.1%)ompared to PPV for discharges of 91.4% (95% CI3.3—95.9%).

OaIt

Table 3 Statistical analysis of admission and discharge predictio

Subjects Prediction Sensitivity (95% CI) % Spec

ED consultants Admission 91 (82—96) 69.8Discharge 69.8 (62—76.7) 91 (8

ED registrars Admission 87.5 (76.3—94.1) 64.4Discharge 64.4 (55.5—72.4) 87.5

Triage nurses Admission 76.7 (70—82.3) 73.9Discharge 73.9 (70.2—77.3) 76.7

193 (24.3%) 602 (75.7%) 795 (100%)

Among triage nurses, the proportion of true positives andalse positives for admission was 48.5% and 51.5% respec-ively. The proportion of true positives and false positives forischarge was 90.8% and 9.2% respectively. PPV for admis-ion was 48.5% (95% CI 42.8—54.3%) compared to PPV forischarges of 90.8% (95% CI 87.8—93.2%).

iscussion

ur study results demonstrate that ED specialist, registrarsnd triage have a very high PPV for discharge predictions.f triage nurses’ predictions of discharged patients wereo be implemented and patients streamed accordingly in

ns made by consultants, registrars and triage nurses.

ificity (95% CI)% PPV (95% CI) % NPV (95% CI) %

(62—76.7) 59.7 (50.3—68.4) 94.1 (87.7—97.4%)1.8—96) 94.1 (87.7—97.4) 59.7 (50.3—68.4)

(55.5—72.4) 54.4 (44.3—64.1) 91.4 (83.3—95.9) (76.3—94.1) 91.4 (83.3—95.9) 54.4 (44.3—64.1) (70.1—77.3) 48.5 (42.8—54.3) 90.8 (87.8—93.2) (70—82.3) 90.8 (87.8—93.2) 48.5 (42.8—54.3)

aetsg

maSpswgmtwtwwwpfToscc

nEplaa

C

TsddniTdbwpc

P

NT

The EPOD study

the department, approximately 7—12% (PPV 87.8—93.2%)of potentially admitted patients would be allocated toobservation wards or into the discharge stream of thepatients. When ED specialists made these streaming deci-sions, approximately 3—12% (PPV 87.7—97.4%) of admittedpatients would be allocated in the discharge stream. Thislevel of discharge failure through the short-stay unit isstill below our key performance indicator (KPI) threshold of<15%.

We also noticed another important finding in this study.ED consultants, ED registrars and triage nurses have PPVsfor admission of 59.7%, 54.4% and 48.5% respectively. Thismeans triage nurses, ED registrars and ED consultants withequal ability can predict patients who are likely to be admit-ted and therefore they can fast track these patients into alocation where they may be evaluated quickly.

Admission rates from ED presentations fluctuate between35% and 40% (including 3—5% short-stay admissions) at West-mead hospital. This group of patients awaits placement intoinpatient hospital beds. At the time of hospital overcrowd-ing and access block, these patients occupy ED treatmentspaces for increasing lengths of time creating impedimentsto smooth patient-flow through the ED. A major challengeduring these periods is to maintain ongoing flow through thedepartment by expediting the process of admission-streamand discharge-stream patients in the department. Whereasadmission-stream patient can be diverted to a dedicatedarea in the ED to have early treatment initiated, discharge-stream patients can be streamlined to a certain extent inorder to maintain a flow within the department.

Overcrowding and prolonged length of stay afflict mostbusy EDs around the world. Whilst there is a lot of literaturesurrounding interventions to tackle access block, most ofthese involve hospital-wide and system-wide measures.16,17

EDs need to continue functioning and maintain intradepartmental patient-flow to meet the ever-increasing EDpresentations. Streaming as a potential solution for improv-ing patient-flow through ED is not a new concept.9,10

Streaming can occur based on resource allocation i.e.complexity7 or predicted disposition outcomes.10 Streamingbased on disposition outcome aims to reduce the impact ofaccess block on the discharge-stream patients and thus helpED maintain patient-flow for the majority of their patients.18

Most tertiary ED now have streamed models of caree.g. resuscitation, acute care, fast-track or walk-in clinicsand/or short-stay units. The streaming of patients to thesemodels of care can occur during various stages of the patientjourney through the ED. In this study, we aimed to determinewhether streaming decisions could be made effectively atearly stages of the patient journey and thus improve effi-ciency within the department.

Patients admitted to the short-stay unit should predomi-nantly be suitable for discharge within 24 h. Inappropriatelyplacing admission-stream patients in a discharge stream cancause patient-flow problems. Thus, most admission crite-ria for short-stay units suggest informed decision makingby senior medical staff in well-differentiated patients. Ourstudy suggests that triage nurses, ED consultants and ED reg-

istrars can make these decisions for the discharge streamearly in the course of their journey through ED with a highlevel of certainty (PPV 87.8—97.4%). Our findings also sug-gest that triage nurses, ED registrars and ED consultant can

F

Tn

165

lso make disposition decisions for an admission stream withqual ability (PPV 48.5—59.7%). EDs can choose to utiliseriage nurses, ED consultants or ED registrars as the mostuitable personnel to make this early streaming decision toain maximum benefit for the department.

Our study has some limitations. Whilst triage nurses andedical officers were asked to fill out prediction forms for

ll patients with ATS categories 3—5 and reviewed in theAFE-T zone, not all patients meeting the criteria had theredictions filled out. This could raise the possibility of aelection bias with a probable suggestion that predictionsere only made in patients where the outcomes were sug-estive at triage. Another significant limitation was that theedical officers making the predictions were sometimes also

hen making the clinical decisions for the patients. Thisould lead to increased correlation between the predic-

ion and the result. The end-points of the study (Table 1)ere designed to reduce the risk of this bias. Patientsho were admitted but then discharged by inpatient teamsithin 24 h were considered as failed predictions. Similarly,atients who were placed in short-stay unit and then stayedor longer than 24 h were considered as failed predictions.hese outcomes were beyond the control of the medicalfficers who were making the initial prediction and deci-ion. We did not contact patients or other health services toheck for representations beyond our hospital, which wouldonstitute a failed discharge.

Another limitation of this study was that whilst triageurses made predictions for all patients presenting to triage,D consultants and ED registrars potentially missed makingredictions on patients directly streamed from triage to theow-acuity urgent care centre. This is reflected in the resultss triage nurses made more predictions on ATS categories 4nd 5 patients than ED consultants and ED registrars.

onclusion

his study has demonstrated that triage nurses, ED con-ultants and ED registrars are able to predict a patient’sischarge disposition at triage with high levels of confi-ence. The findings of this study also suggest that triageurses, ED consultants, and ED registrars with equal abil-ty can predict the patients who are likely to be admitted.his data may be used to develop specific admission andischarge streams based on early decision-making in EDsy triage nurses, ED registrars or ED consultants. Thisould facilitate patient-flows in the departments aidingatient-flow during periods of access block and ED over-rowding.

rovenance and conflict of interest

one of the authors have any conflict of interest in this study.his paper was not commissioned.

unding

here was no funding required for this study as there waso change in current practice in ED.

1

E

WEtHt

A

Aceh

R

1

1

1

1

1

1

1

1

66

thical approval statement

estern Sydney Local Health District Human Researchthics Committee approval was obtained for a prospec-ive observational study with reference numberREC2012/6/6.1(3541)AU RED LNR/12/WMEAD200 forhis study.

cknowledgement

uthors acknowledge all Westmead ED medical, nursing andlerical staff that participated in this study. We acknowl-dge Karen Byth, Biostatistician at Westmead Hospital forer statistical guidance.

eferences

1. Australian hospital statistics 2010—11. Canberra: AIHW: Aus-tralian Institute of Health and Welfare; 2012.

2. Cooke MW, Higgins J, Kidd P. Use of emergency observation andassessment wards: a systematic literature review. Emerg Med J2003;20(2):138—42.

3. McNeill GB, Brand C, Clark K, Jenkins G, Scott I, Thomp-son C, et al. Optimizing care for acute medical patients: theAustralasian Medical Assessment Unit Survey. Intern Med J2011;41(1a):19—26.

4. Ong BS, Van Nguyen H, Ilyas M, Boyatzis I, Ngian VJ. Med-ical assessment units and the older patient: a retrospectivecase—control study. Aust Health Rev 2012;36(3):331—5.

5. Farrohknia N, Castren M, Ehrenberg A, Lind L, Oredsson S, Jon-sson H, et al. Emergency department triage scales and theircomponents: a systematic review of the scientific evidence.Scand J Trauma Resusc Emerg Med 2011;19:42.

6. Forero R, Nugus P. Australasian College for Emergency Medicine

(ACEM) literature review on the Australasian triage scale (ATS).Sydney: University of New South Wales — Simpson Centre forHealth Services Research, South Western Sydney Clinical School,Centre for Clinical Governance Research; 2012.

1

M.R. Vaghasiya et al.

7. Leraci S, Digiusto E, Sonntag P, Dann L, Fox D. Streaming by casecomplexity: evaluation of a model for emergency departmentfast track. Emerg Med Australas 2008;20(3):241—9.

8. Rodi SW, Grau MV, Orsini CM. Evaluation of a fast trackunit: alignment of resources and demand results in improvedsatisfaction and decreased length of stay for emergencydepartment patients. Qual Manag Health Care 2006;15(3):163—70.

9. Kelly AM, Bryant M, Cox L, Jolley D. Improving emergencydepartment efficiency by patient streaming to outcomes-basedteams. Aust Health Rev 2007;31(1):16—21.

0. King DL, Ben-Tovim DI, Bassham J. Redesigning emergencydepartment patient flows: application of lean thinking to healthcare. Emerg Med Australas 2006;18(4):391—7.

1. Shetty A, Gunja N, Byth K, Vukasovic M. Senior Streaming Assess-ment Further Evaluation after Triage zone: a novel model ofcare encompassing various emergency department throughputmeasures. Emerg Med Australas 2012:374—82.

2. Levine SD, Colwell CB, Pons PT, Gravitz C, Haukoos JS, McVaneyKE. How well do paramedics predict admission to the hospital?A prospective study. J Emerg Med 2006;31(1):1—5.

3. Holdgate A, Morris J, Fry M, Zecevic M. Accuracy of triagenurses in predicting patient disposition. Emerg Med Australas2007;19(4):341—5.

4. Price TG, Hooker EA, Neubauer J. Prehospital providerprediction of emergency department disposition: implica-tions for selective diversion. Prehosp Emerg Care 2005;9(3):322—5.

5. Beardsell I, Robinson S. Can emergency department nurses per-forming triage predict the need for admission? Emerg Med J2011;28(11):959—62.

6. Cameron PA, Joseph AP, McCarthy SM. Access block can be man-aged. Med J Aust 2009;190(7):364—8.

7. Bernstein SL, Asplin BR. Emergency department crowd-ing: old problem, new solutions. Emerg Med Clin N Am2006;24(4):821—37.

8. Saghafian S, Hopp WJ, Van Oyen MP, Desmond JS, KronickSL. Patient streaming as a mechanism for improving respon-siveness in emergency departments. Oper Res 2012;60(March(5)):1080—97. Ross School of Business Paper #1162.