emergency department interventions to reduce ed ... · non–emergency department interventions to...
TRANSCRIPT
PROGRESSIVE CLINICAL PRACTICE
Non–Emergency Department Interventionsto Reduce ED Utilization: A SystematicReviewSofie Rahman Morgan, MD, MBA, Anna Marie Chang, MD, MSCE, Mahfood Alqatari, MD, and JesseM. Pines, MD, MBA, MSCE
AbstractObjectives: Recent health policy changes have focused efforts on reducing emergency department (ED)visits as a way to reduce costs and improve quality of care. This was a systematic review of interventionsbased outside the ED aimed at reducing ED use.
Methods: This study was designed as a systematic review. We reviewed the literature on interventions infive categories: patient education, creation of additional non-ED capacity, managed care, prehospitaldiversion, and patient financial incentives. Studies written in English, with interventions administeredoutside of the ED, and a comparison group where ED use was an outcome, were included. Twoindependent reviewers screened search results using MEDLINE, Cochrane, OAIster, or Scopus. Thefollowing data were abstracted from included studies: type of intervention, study design, population,details of intervention, effect on ED use, effect on non-ED health care use, and other health and financialoutcomes. Quality of individual articles was assessed using Grading of Recommendations Assessment,Development, and Evaluation (GRADE) guidelines.
Results: Of 39 included studies, 34 were observational and five were randomized controlled trials. Two offive studies on patient education found reductions in ED use ranging from 21% to 80%. Out of 10 studiesof additional non-ED capacity, four showed decreases of 9% to 54%, and one a 21% increase. Bothstudies on prehospital diversion found reductions of 3% to 7%. Of 12 studies on managed care, 10 haddecreases ranging from 1% to 46%. Nine out of 10 studies on patient financial incentives founddecreases of 3% to 50%, and one a 34% increase. Nineteen studies reported effect on non-ED use withmixed results. Seventeen studies included data on health outcomes, but 13 of these only included data onhospitalizations rather than morbidity and mortality. Seven studies included data on cost outcomes.According to the GRADE guidelines, all studies had at least some risk of bias, with four moderatequality, one low quality, and 34 very low quality studies.
Conclusions: Many studies have explored interventions based outside the ED to reduce ED use invarious populations, with mixed evidence. Approximately two-thirds identified here showed reductionsin ED use. The interventions with the greatest number of studies showing reductions in ED use includepatient financial incentives and managed care, while the greatest magnitude of reductions were found inpatient education. These findings have implications for insurers and policymakers seeking to reduce EDuse.
ACADEMIC EMERGENCY MEDICINE 2013; 20:969–985 © 2013 by the Society for Academic EmergencyMedicine
From the Department of Emergency Medicine, Emory University (SRM), Atlanta, GA; the Department of Emergency Medicine,Oregon Health and Science University (AMC), Portland, OR; the Department of Emergency Medicine, George Washington Univer-sity (MA), Washington, DC; and the Departments of Emergency Medicine and Health Policy, George Washington University Hos-pital (JMP), Washington, DC.Received January 25, 2013; revision received April 14, 2013; accepted April 16, 2013.Presented at the American College of Emergency Physicians Scientific Assembly, Denver, CO, October 2012.Dr. Chang is generally supported by Award Number 1K12HL108974-01 from the National Heart, Lung, and Blood Institute. Dr.Pines is an associate editor for this journal, but had no involvement with the review or publication decision for this manuscript.Supervising Editor: Shahriar Zehtabchi, MD.Address for correspondence and reprints: Sofie Rahman Morgan, MD, MBA; e-mail: [email protected] related commentary appears on page 1062.
doi: 10.1111/acem.12219 PII ISSN 1069-6563583 969© 2013 by the Society for Academic Emergency Medicine ISSN 1069-6563 969
Growing health care costs in the United Stateshave made patients, providers, and payersexamine the value of services delivered.1 Con-
cepts such as accountable care organizations and medi-cal homes are gaining momentum with the goal oflimiting avoidable, redundant, ineffective, or harmfultreatments in favor of expanding effective care, accessto care, and care coordination.
Many programs aimed at improving efficiency focuson the use of hospital-based emergency departments(ED) for care. EDs care for critically ill patients andacute unscheduled conditions and serve as a safety netfor those with limited access to health care due to insur-ance status, the timely availability of clinic-based physi-cians, and the need for care outside of traditionalbusiness hours.2 The focus on the ED as a place toimprove efficiency stems from observations that EDcare for low-acuity conditions results in higher chargesthan for similar diagnoses seen in other settings.3 Inaddition, an ED visit may be a marker of a potentiallyavoidable injury or illness that could have been pre-vented with better primary care, patient education, orenhanced public health measures.
Studies have examined the effect of interventions toreduce ED use that are performed outside the ED, suchas patient education, improved clinic access, care coor-dination, patient-centered care, and others. While ED-based interventions also exist, they are fundamentallydifferent because of their location and their focus (e.g.,follow-up vs. prevention). Our group recently conducteda systematic review of ED-based care coordinationinterventions.4 Therefore, this review focuses specifi-cally on interventions based outside of the ED lookingat systems-level changes, rather than ED-specificchanges. Prior reviews of aggregated non-ED interven-tions have either focused only on one type of interven-tion or excluded some subsets of visits such as pediatricpatients or categories of intervention such as prehospi-tal diversion.5–7 To our knowledge, there has been nobroad-based inclusive review of the comparative effec-tiveness of the myriad interventions tested to reduce EDuse. The goal of this investigation was to review the evi-dence on the effectiveness of interventions based out-side of the ED aimed at reducing ED use and,ultimately, to explore themes about which interventionsmay be most effective, along with any undesired conse-quences.
METHODS
Study DesignWe systematically reviewed the literature on the effec-tiveness of non-ED interventions aimed at reducing EDuse. Non-ED interventions were defined as those imple-mented outside of an ED or hospital (e.g., insurance-based, outpatient clinic-based). No human subjects ormedical records were reviewed as a part of this studyso institutional review board approval was notrequired.
The databases MEDLINE, Cochrane, OAIster, andScopus were examined from 1966 to the present. Key-words used included emergency department, emer-gency medical services, utilization, demand, patient
education, primary care, capacity, extended hours,advanced access, telephone triage, general practitioner,care coordination, copayment, and payment reform(MEDLINE search terms in Appendix A). Searches werelimited to English language publications. Because of thevariety in the interventions and outcome measurementsin the results, we performed a qualitative rather thanquantitative systematic review.
Data Collection and ProcessingTwo independent reviewers (MA, SRM) screened searchresults and excluded those with titles that did not fitinclusion criteria (next section). The two reviewers thenscreened the abstracts of the remaining citations, againexcluding those that did not fit inclusion criteria. Theremaining articles underwent full-text review to excludeany remaining studies that did not fit inclusion criteria.Intrarater reliability was measured with a 10% sampleof citations, resulting in a kappa of 0.92. Each articlewith conflicting opinion from reviewers was discussedwith a third reviewer (JMP) for a final resolution.
The following data were abstracted from all eligiblestudies: type of intervention, study design, population,details of intervention, effect on ED use, effect on non-ED health care use, and other health and financial out-comes. Reviewers used a standard format to abstractdata; this format mirrors the categories in the tablespresented here. We attempted to standardize resultsacross studies. In studies with data available for abso-lute number of ED visits before and after the interven-tion was implemented (as opposed to, for example,number of visits per person, etc.), we calculated apercentage reduction of ED visits. When visit numberswere reported, the difference between number of visitsbefore and after the intervention was divided by thenumber of ED visits prior to intervention to standardizethe comparison of study results.
We followed guidelines created in the PreferredReporting Items for Systematic Reviews and Meta-Anal-yses (PRISMA) statement to create a four-phase flowdiagram (Data Supplement S1, available as supportinginformation in the online version of this paper) showingthe number of records included and excluded at eachphase.8
Inclusion and Exclusion CriteriaStudies were included if they had interventions adminis-tered outside of the ED, had a comparison group whereED use was an outcome, and were in English. Whilestudies in other languages were excluded, studies wereincluded regardless of country and health care system ifpublished in English. Five categories of interventionswere included: 1) patient education on medical condi-tions and appropriate medical care use for low-acuityconditions, 2) creation of additional capacity in non-EDsettings (e.g., expanded hours or same-day access), 3)managed care (e.g., primary care physician capitationor gatekeeping), 4) prehospital diversion, and 5) patientfinancial incentives (e.g., copayments or deductibles).Two other interventions, telephone triage and casemanagement, were initially searched for but becauserecent systematic reviews have compiled the results ofthose topics; these were excluded.5,6
970 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
Studies with ED-based interventions were excluded,as were studies without measurable objective outcomes.Studies with outcomes assessed through patient or pro-vider subjective surveys were also excluded.
Quality AssessmentQuality assessment was done using the portion of theGrading of Recommendations Assessment, Develop-ment, and Evaluation (GRADE) criteria aimed at assess-ing the risk of bias in individual articles.9 Because of theheterogeneity in study designs, other components of theGRADE criteria, including the formal overall evaluationof the body of literature, were not used. Additionally,studies with risk-adjusted results and where significancewas measured are reported.
RESULTS
Literature SearchThe search yielded 793 titles. After removing duplicatesand exclusions based on title or abstract, 62 studiesremained and underwent full-text review (Data Supple-ment S1). An additional 19 references were identifiedand also underwent full-text review, which resulted in atotal number of included studies of 39 that ranged frompublication dates of 1986 to 2011. The number ofincluded studies per category is as follows: patient edu-cation on medical conditions and health care use, fivestudies10–14; creation of additional non-ED capacity, 10studies15–24; prehospital diversion of low-acuity patients,two studies25,26; managed care, 12 studies27–38; andpatient financial incentives, 10 studies.39–47
Description of Included StudiesPatient Education on Medical Conditions Health CareUse. Two out of five studies found significant reduc-tions in the use of the ED after interventions, withreductions ranging from 21% to 80% (Table 1).10–14 Allstudies were based in the United States. Interventionsincluded use of booklets or in-person educational ses-sions. Both pediatric and adult patients are included.Three studies reported data on non-ED use with onefinding 0.03 fewer clinic visits per person.10–12 Threearticles reported health outcomes and no significantadverse events were noted.11–13 No studies reportedrisk-adjusted data.
Capacity Increase in Non-ED Settings. Of 10 studies,three examined interventions that expanded capacitythrough new community clinics, while the remainderinvolved existing physician practices expandingappointments and/or hours of care. Four studies foundsignificant decreases in the use of the ED after increasesin non-ED capacity, with reductions ranging from 9%to 54%, while five were nonsignificant and one foundan increase of 21% (Table 2).15–24 Four studies werebased in the United States and the remaining six inCanada or Europe. Regarding effect on non-ED use,five studies reported data with four showing increasesin non-ED use ranging from 1% to 102%.16,17,20–23 Ofthese, one article reported that while there was anincrease in primary care use, there was a concurrentdecrease in urgent care use.20 Two articles reported
health outcomes.20,21 Three studies reported cost datashowing 10% to 20% savings with the interven-tion.15,18,20 Two studies risk-adjusted data.20,21
Prehospital Diversion of Low-acuity Patients. Bothstudies examining the effects of emergency medical ser-vices (EMS) diversion of low-acuity patients away fromthe ED found significant decreases in the use of the EDafter interventions, with reductions ranging from 3% to7% (Table 3).25,26 One study was conducted in the Uni-ted States, and the other, in the United Kingdom. Oneintervention involved EMS offering either home orclinic care to low-acuity patients.25 The other involvedtransportation of such patients to clinic care withouthome care as an option.26 Regarding effect on non-EDuse, both studies found increases in use of other caresettings. No studies directly addressed other health orcost outcomes. No data were risk adjusted.
Managed Care. Of the 12 studies examining theeffects of managed care on ED use, six had interven-tions with capitated payment of primary care physician,five had a requirement of primary care physicianapproval or gatekeeping, and one was a hybrid of thesetwo (Table 4).27–38 Ten studies were based in the UnitedStates, one in Canada, and one in Ireland. The majorityof U.S. studies were in Medicaid populations andincluded pediatric patients. Overall, nine studies (sixwith capitation and four with gatekeeping as interven-tions) found significant decreases in the use of the EDafter managed care interventions, with reductions rang-ing from 1% to 46%, while two did not find any signifi-cant difference.27–31,33–38 The final study found mixedresults with no change in ED use when comparing phy-sicians pre- and postcapitation, but with an increase inED use among physicians compensated through capita-tion versus fee-for-service.32
Regarding the effect on non-ED use, six studies didreport data, with only four reporting significance andmixed results.28,30,32,35,37,38 Six articles included healthoutcome data, five in the form of effect on hospitaliza-tions and one on morbidity indices; results were mixedand did not always assess for significance.27,30,32,34,37,38
Two studies reported cost data with both showingdecreases with capitation.28,30 No studies reported risk-adjusted data.
Patient Financial Incentives. Of the 10 studies usingcosts to influence patients to use certain sites for care,or to use care efficiently, nine studies found significantdecreases in the use of the ED after implementation ofthe intervention, with reductions ranging from 3% to50% (Table 5).39–47 The remaining study found a signifi-cant relative increase of 34% in ED visits.48 All studiestook place in the United States. The intervention inseven studies was the requirement for patient copay-ment or coinsurance, and in three it was the implemen-tation of a high deductible. Half of the studies were inMedicaid populatons, with the majority of those single-state interventions, while the others involved commer-cial insurers.
Regarding effect on non-ED use, two studies didreport data with one showing no change in urgent care,
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 971
Table
1PatientEduca
tiononHealthCare
Use
Author/Year
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:Booklet
Glavanetal.
(1998)10
Pre–p
ost
interventional
study
UnitedStates
6months
Air
Forcebase
Interventiongroup:
1,555su
bjects
Controlgroup:
972su
bjects
Subjectsin
intervention
groupwere
given
ase
lf-care
bookwith
aninform
ation
sess
iononhow
touse
thebook
Interventiongroup:
reductionin
ED
use
rate
perperson
from
0.353to
0.279(p
=0.02)
Controlgroup:
increase
inED
use
rate
perpersonfrom
0.386to
0.421
Interventiongroup:
reduction
inclinic
use
rate
per
personfrom
1.072to
1.038
Controlgroup:increase
inclinic
visitsrate
perperson
from
1.040to
1.168
Notreported
Rectoretal.
(1999)14
Randomized,parallel
groupstudy
UnitedStates
January–J
uly
1998
TwourbanMedicaid
health
plans(PlanA
andB)
Interventiongroup:
house
holds
senteduca
tionalbroch
ure
Controlgroup:
house
holdsnot
senteduca
tionalbroch
ure
House
holdswere
randomizedto
rece
ive
abooklet,First
Look,
aboutca
reof
commonnonurgent
conditions
Nonsignifica
nt
reductionsin
ED
use
of1.1%
(PlanA,95%
CI=–3
.1%
to0.8%)
and–1
.2%
(PlanB,95%
CI=
–4.1%
to1.4%)
Nodifference
ineitherplan
inphysicianoffice
visits
Intervention:TrainingSess
ions
McW
illiams
etal.
(2008)11
Retrosp
ective
cohort-control
study,difference
-in-
difference
model
UnitedStates
12months
Primary
care
practicesat
aca
demic
healthce
nter
Interventiongroup:191
patients
inpractice
Controlgroups:
•168patients
at
samepractice
from
previousyear
•2co
hortsfrom
same
yearatprimary
care
siteswithout
intervention,
133and126patients
Nursesprovided
standardizededuca
tion
alongwith
presc
riptionfor
pain
relievingeardrops
Parents
remindedof
24-hourmedical
advicetelephoneaccess
Reductionin
ED
use
forearpain
by80.3%
(p=0.009)
atinterventionsite.
Concu
rrentco
ntrol
site
had25%
nonsignifica
nt
increase
inED
use
Nonsignifica
ntreductionsin
urgentca
rece
nteruse
by40.3%
(p=0.33),and
primary
care
centeruse
by27.8%
(p=0.14)at
interventionsite.In
controlsite,28%
reduction
inurgentca
reuse
,anda4%
reductionin
primary
care
use
3-yearmedicalreco
rdreview
ofallch
ildren
intheintervention
grouprevealedno
episodesofmastoiditis
Rettig
etal.
(1986)12
Randomized
controlledtrial
UnitedStates
Oneyear
Multiple
homehealth
nursingagencies
Interventiongroup:180
diabeticpatients
Controlgroup:193
diabeticpatients
Diabeticpatients
inthe
interventiongroup
underw
enthome
teach
ingse
ssions
ondiabetichealth
problems
Nonsignifica
nt
reductionin
ED
visitsin
intervention
vs.
control
group(0.06vs.
0.08)
(nonsignifica
nt)
Nonsignifica
ntincrease
inprimary
care
visitswithin
6monthsby
interventionvs.
controlgroup
(3.11vs.
2.78)
Nonsignifica
ntch
anges
inhosp
italizationrates/
1,000members/yearin
interventiongroupvs.
controlgroup:
972 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
pediatric office, adult office, and ambulatory care visits,but one showed an increase in hospital outpatientdepartment use.42,43 Six studies reported health out-comes in the form of ED visits resulting in hospitaliza-tion, which decreased or were unchanged.39–42,44,47 Ofthese six, two also reported effects on morbidity, oneshowing no change and one showing a decrease.39,42
Three studies reported cost data with mixed results.43–45
Five studies risk-adjusted data.39,42,45,46,48
Quality AssessmentQuality assessment for risk of bias in individual articlescan be found in Table 6 and Data Supplement S2 (avail-able as supporting information in the online version ofthis paper). All 34 observational studies were of verylow quality according to GRADE guidelines, in whichobservational studies are at best low quality and anyserious risk of bias lowers the quality to very low. Fourof the randomized trials were of moderate quality andone of low quality.
DISCUSSION
In 2003, Asplin et al.2 identified input, throughput, andoutput factors associated with crowding in the ED. Inthis paper, we explore primarily the effect of how sys-tems outside the ED can influence demand for ED careand focused entirely on “input.” This expands on previ-ous work where we conducted a systematic review ofcare coordination within the ED.4 With respect to non-ED interventions, we found there have been many stud-ies exploring non-ED interventions to reduce ED use invarious populations across more than two decades withmixed evidence. Over two-thirds (27 of 39, 69%) of theidentified studies in the five categories of interventionincluded showed reductions in ED use, with reductionsranging from 1% to 80%. Nearly all of the studies wereobservational, many coinciding with systemic changeswithin the region, and only five out of the 39 were ran-domized trials. Only seven of 39 studies (17.9%)reported risk-adjusted data.
The areas with the largest number of studies showingreductions in ED use include patient financial incentivesand managed care interventions, with nine of 10 (90%)and 10 of 12 (83%) of those studies showing reductions,respectively. By contrast, less than half (four of 10) ofthe studies on increasing capacity found reductions inED use, and one even found an increase, suggestingthat adding capacity could potentially have the oppositeeffect. This may be due to issues with supply-induceddemand. Findings on capacity should be interpretedcarefully, however, as the majority of articles on thistopic are international, and several take place in single-payer health care systems. The area showing the largestmagnitude of reduction is patient education, with amaximum of an 80% reduction in one study.
We hope to build on the growing body of literaturethat addresses the overall aim to reduce ED use. Priorreviews exist on two topics not included here, telephonetriage and case management. For telephone triage, aCochrane review of nine studies included seven studiesthat examined the effect on ED use.5 Of these, sixshowed no difference, and one showed an increase inT
able
1Continued
Author/Year
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
•Non–d
iabetesrelated
(544.4
vs.
440.4)
•Nonpreventable
diabetesrelated
(66.7
vs.
82.9)
•Preventable
diabetes
related(94.4
vs.
41.5)
Sch
onlau
etal.
(2005)13
Pre–p
ost
interventional
study
UnitedStates
Primary
healthca
rece
nters
recruited
throughBreakthrough
SeriesCollaborative
Interventiongroup:three
siteswith109patients
Controlgroup:tw
ositeswith76patients
Asthmaticpatients
at
interventionsites
underw
ent
educa
tionalse
ssions
Nonsignifica
nt
difference
sin
acu
teca
revisitsbetw
een
groups(1.72vs.
0.92average
visits;
p=0.08)
Notreported
Nosignifica
ntdifference
inquality
oflife,number
ofbeddays,
andacu
teca
rese
rviceuse
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 973
Table
2Capacity
Increase
inNon-ED
Settings
Author(Year)
Design;Country;
DurationofStudy
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:AdditionalClinic(s)
Chalderetal.
(2003)2
2Matchedtime-series
analysis
England
Oneyearbefore
andafter
NHS
developedwalk-ince
nters
withdrop-inse
rvice
Interventiongroup:10towns
withwalk-ince
nter
Controlgroup:10similar
size
townsin
same
regionwithoutwalk-ince
nters
Walk-ince
nters
provided
nurse-led,
drop-inse
rvicewith
broadhours
ofoperation
Decrease
of173.3
visits/month
vs.
decrease
ofthreevisits/
month
inco
ntrolgroup
(p=0.11)
Before
andafterincrease
of3.9
visitsto
primary
care
facility
vs.
23.7
visitsin
controlgroup(p
=0.25)
None
Hsu
etal.
(2003)2
4Before-and-after
obse
rvationalstudy
withco
ntrolgroup
England
6monthsbefore
andafter
NHS
developedwalk-ince
nters
withdrop-inse
rvice
Interventiongroup:onetown
withwalk-ince
nter
Controlgroup:onenearby
townin
thesa
me
regionwithoutwalk-ince
nter
Walk-ince
nters
providednurse-led,
drop-inse
rvicewith
broadhours
ofoperation
ED
visitsincrease
dby10%
(adjustedRR
=1.10;95%
CI=1.00to
1.21)
0.10feweremergency
consu
ltationsperday
(95%
CI=–3
.75to
3.55)
None
O’Kellyetal.
(2010)1
6Retrosp
ectivestudy
Ireland
1999–2
007
Single,largeED
andout-of-hours
generalpractice
emergency
serviceclinic
associatedwithsa
mehosp
ital
Interventiongroup:Dubdocpatients
Controlgroup:low-triage-level
ED
patients
“Dubdoc”
isan
out-of-hours
general
practiceemergency
servicesclinic
Significa
ntdecrease
low-triage-levelED
visitsduringDubdoc
hours
(54%
decrease
inlow-acu
ityED
visitsvs.
52%
decrease
outside
Dubdochours;p<0.033)
Increase
of102%
in“Dubdoc”
visits(3,810in
1999to
7,696in
2007)
Notreported
Rust
etal.
(2009)1
9Cohort
study
UnitedStates
2003–2
005
Ruralco
untiesin
state
ofGeorgia
Interventiongroup:co
unties
withCHCs(n
=24)
Controlgroup:co
untieswithout
federallyfundedCHCs(n
=93)
CHCsprovideca
reto
theuninsu
red
withoutmedicalhomes
Non-CHC
countieshad
higherratesofED
visits
(RR
=1.21;95%
CI=
1.02to
1.41)
Nonereported
Nonereported
Intervention:Changein
scheduling/hours
Hudecetal.
(2010)1
5Case
-compariso
nstudy
before-and-after
advance
access
implementation
Canada
3months
Fourfamilyphysicianpractices
Interventiongroup:patients
of
generalpractice
withadvance
daccess
booking
Controlgroup:patients
ofthreegeneral
practiceswithtraditionalbooking
Practiceinstituteda
modelin
which
most
appointm
ents
are
same-dayaccess
28%
reductionin
low-acu
ity
ED
visitsin
group
None
7%
revenueincrease
forpractice
Improvedse
lf-reported
patientsa
tisfaction
(p<0.05)
Philipsetal.
(2010)2
3Before-and-after
obse
rvational
studywith
controlgroup
Belgium
2006–2
007
Form
ationofGPCs
Interventiongroup:city
(Turnhout)
that
implementedGPC
Controlgroup:tw
oother
largecitiesthat
did
nothaveGPC
(GhentandAntw
erp)
GPC
reorganized
providers
forthatregion
andce
ntralizedout-of-hours
primary
healthce
nter,
openon
weeke
ndsandholidays
815visitsto
ED
before
and
791visitspost
(no
significa
ntch
ange)
Primary
care
visitsincrease
din
theinterventionregion
(714to
1197visits)
(OR
=1.37;95%
CI=1.20
to1.57),whilestayed
relatively
constantin
controlregion
(734to
850visits)
Solberg
etal.
(2004)2
0Retrosp
ective
pre–p
ost
study
UnitedStates
1999–2
001
Riskadjusted
Large,multispecialtymedicalgroup
Patients
withthreese
lect
chronic
conditions
(heart
disease
,diabetes,
depression)
Interventiongroup:
17,376patients
in2001afterim
plementation
ofopenaccess
Controlgroup:16,099
patients
in1999before
implementationofopenaccess
Fulladvance
daccess
appointm
ents
introduce
d,which
included
standardizationof
schedule
slots
andextravisittime
forclinicians
Nosignifica
ntdifference
inrisk-adjustedproportionof
patients
visitingED
(during
this
timeperiod,ED
visits
alsoincrease
dby7.8%
for
themedicalgroup)
(allp>0.05)
Increase
inprimary
care
visits(allp<0.01)
Significa
ntdecrease
inrisk-adjustedproportion
ofpatients
visitingurgent
care
(allp<0.001)
Totalco
stperpatients
increase
d10%–2
0%
dependingonco
ndition
Significa
ntreductionof
proportionofpatients
with
hosp
italizations(1%)
andlength
ofstay>3days
(4%)forheart
disease
patients
only
(nonsignifica
ntch
anges
forotherco
nditions)
974 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
Table
2Continued
Author(Year)
Design;Country;
DurationofStudy
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Solberg
etal.
(2006)2
1Retrosp
ective
pre–p
ost
study
UnitedStates
1999–2
001
Riskadjusted
Large,multispecialtymedicalgroup
ownedbyhealthplan
Interventiongroup:6,609
patients
with
depressionin
2001after
implementation
ofopenaccess
Controlgroup:7,284patients
withdepression
in1999before
implementation
ofopenaccess
Advance
daccess
appointm
ents
instituted(third
next-dayavailable
appointm
ent
reduce
dfrom
19.4
to4.5
days)
Nosignifica
ntch
ange
inED
visits
1%
increase
inprimary
care
visits(p
<0.01)
2%
increase
inhosp
italizations(p
<0.05)
17.6%
reductionin
proportion
ofpatients
withnofollow-up
afterstartingnew
medication
(p=0.001)
Majority
ofvisitswithone
physician(continuityofca
re)
increase
dby6.7%
(p<0.001)
vanUden
etal.
(2004)1
7
Retrosp
ective
pre–p
ost
cohort
study
Lim
burg,
theNetherlands
Regionalgeneralpractitioner
cooperatives
Interventiongroup:12,319
patientco
ntacts
afterim
plementationofGPCs
Controlgroup:11,781
patientco
ntacts
before
implementationofGPCs
Largeco
operativesof
generalpractitioners
toofferoutofhours
primary
care,alsoact
as
gateke
eperforED
visits
9%
decrease
inED
visits
afterhours
13.7%
abso
lute
reduction
numberofse
lf-referrals
toED
10%
increase
inprimary
care
visitsafterhours
4.6%
increase
inoverall
patientco
ntactsafterhours
3.6%
shiftin
use
from
ED
toprimary
care
(p<0.001)
None
Wangetal.
(2005)1
8Pre–p
ost
intervention
studywith
compariso
ngroup
UnitedStates
12months
Large,private,primary
care
pediatric
practicewithMedicaid
patients
Interventiongroup:17,382
childrenin
the
enhance
daccess
program
Controlgroup:26,066
Medicaid-eligible
childrenwhorece
ived
servicesfrom
other
loca
lco
mmunityprimary
care
providers
Increase
dca
reco
ordination,ca
semanagement,expanded
after-hours
clinicsandwalk
inhours
atclinic
20%
reductionin
ED
utilizationforthe
intervention
group(p
=0.007)
Cost
permember
permonth
was
$8.53wasfoundin
the
controlgroup
and$7.17in
intervention
group,
thusaco
mparative
savingsof16%
CHCs=co
mmunityhealthce
nters;GPCs=grouppractitionerco
operatives;
RR
=rate
ratio.
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 975
ED use. A review of interventions targeted at frequentED use included 11 studies, of which the intervention inseven was case management.6 The results were mixed,with the majority of observational trials showing reduc-tions in ED use, but one showing an increase, and onlyone out of three randomized trials showing a reduction.
More recently, Flores-Mateo et al.7 examined inter-ventions that increased the supply of non-ED servicesand those that reduced demand. Similar to our findings,the authors concluded that increasing the supply of pri-mary care physicians and cost sharing with patientswere effective in reducing ED use, but increased out-of-hours primary care was not. However, there were dif-ferences in our methods, especially inclusion criteria.We included different types of studies, specifically pedi-atric studies, physician capitation, advanced accessscheduling, and prehospital diversion and excludedED-based studies, telephone triage, those with surveydata, and those without controls. As a result, theauthors of the prior review found that educational inter-ventions did not appear to reduce ED use, while wefound the opposite.
We also reviewed the articles for the effect on utiliza-tion of other health care settings. About half of the arti-cles reported this data, although the results werepresented in different ways (primary care visits, urgentcare visits, specialist visits, etc.) and varied significantlyby category. Two of five (40%) of the education studiesthat examined non-ED use found decreases in use,while the others found no significant effect, suggestingthat perhaps education may reduce use in general,rather than the possibility of induced demand by addingcapacity. Both of the studies on prehospital diversionand five of 10 (50%) of those on capacity increasesfound increases in non-ED use, which may represent atype of substitution effect in which patients do notreduce their overall consumption of health care, butinstead shift it to other settings. Again, many of thesestudies took place in single-payer systems wherepatients may have more access to primary care settings,whereas in the United States, access could limit thisshift. Only two of 10 (20%) studies on patient financialincentives examined non-ED use, and these showedreductions. Often, though, copayments were imple-mented both inside and outside the ED, so it is difficultto tease apart direct and indirect effects on ED use.Interventions involving managed care showed mixedresults with some increasing and some decreasing non-ED use.
Seventeen of the 29 studies (43.5%) reported adverseevents, and the outcomes used varied widely, althoughno significant adverse events were attributed to inter-ventions. The health outcome that was generally mea-sured was hospitalization, and this is really a secondaryoutcome, with morbidity and mortality being the pri-mary health outcome of interest. Similarly, only eight ofthe 39 articles (20.5%) reported any cost data, whichwas both limited and mixed. Some of the interventionsmay have ethical questions. For one, those with finan-cial implications for patients may result in patientsavoiding needed care and could result in worse healthoutcomes, and in the long run, possibly increased coststo the health care system. Another example is systemsT
able
3Prehosp
italDiversionofLow-acu
ityPatients
Author(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
Sch
aeferetal.
(2002)25
Cohort
studywithmatched
historica
lco
ntrols
UnitedStates
August
2000–J
anuary
2001
TwoEMS
Interventiongroup:
1,016patients
Control(historica
l)group:2,617patients
Low-acu
itypatients
intheintervention
groupwere
offered
analternate
care
sources(clinic
orhome
care)ratherthan
ED
use
bytheEMSstaff
There
was7%
fewerED
use
bytheinterventiongroup
comparedto
theco
ntrolgroup
(44.6%
vs.
51.8%)(p
=0.001)
Clinic
use
:there
was3.5%
more
clinic
use
bytheintervention
groupco
mparedto
theco
ntrol
group(8%
vs.
4.5%)(p
=0.001)
Notransp
ort
(homeca
re):there
was3.7%
more
homeca
re(no
transp
ort)bytheintervention
groupco
mparedto
theco
ntrol
group(47.4%
vs.
4.5)(p
=0.043)
Snooks
etal.
(2004)26
Clusterrandomized
controlledtrial
UnitedKingdom
6months
TwoEMS.
Interventiongroup:
409patients
Controlgroup:
425patients
Low-acu
itypatients
intheintervention
groupwere
tran
sportedto
aMIU
bytheEMSstaff
There
was2.8%
fewerED
use
bytheinterventiongroup
comparedto
theco
ntrol
group(74.1%
vs.
76.9%)
MIU
use
:there
was1.3%
more
MIU
use
bythe
interventiongroupco
mpared
totheco
ntrolgroup(10%
vs.
8.7%)
MIU
=minorinjuriesunit.
976 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
Table
4ManagedCare
(CapitationorGateke
eping)
Author(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:Gateke
eping
Badgett
(1986)2
7Before-and-after
study
UnitedStates
3years
1983–1
985
Single,pediatric
ED
Interventiongroup:patients
assignedto
PCPunder
AFDC
recipients,
estim
ated~4
0,000
Controlgroup:sa
megroup
ofpatients,1yearprior
tostart
ofprogram
Citicare
program
(PCPapproval
requiredforED
services),
providedphysicianco
verage
24hours
aday,7daysaweek
TotalED
visits:
Pre:35,704
During:25,543
Post:31,248
TotalED
use
decrease
d23%
during
thestudyperiod
comparedto
periodafterend
ofprogram,for
patients
enrolledin
AFDC
program,
ED
use
decrease
d46%
Notreported
Totalhosp
italadmissionsfrom
ED
(nosignifica
ntch
ange):
Pre:3,545
During:3,555
Post:3,922
Franco
etal.
(1997)3
1
Prosp
ective
cohort
studywith
historica
lco
ntrols
Single,university-base
dpediatric
clinic.
Medicaid
patients.
Interventiongroup:4,766
patients
(July
1–A
ugust
30,1991)
Controlgroup:2,798
patients
(July
1–A
ugist
30,1981)
Program
requiringPCP
approvalforED
use
ED
visitsdeclined
2.4%
(p=0.00005)
Inappropriate
ED
visits
declined33%
(p<0.00001)
Notreported
Notreported
Hurley
etal.
(1989)3
3
Retrosp
ective
cohort
study
FourMedicaid
demonstration
programs
Interventiongroup:AFDC
inprogram
requiring
gateke
eperapproval
Controlgroup:recipients
oftraditionalMedicaid
programs
AssignedPCPwithapproval
requiredforED
services
Proportionofpatients
withone
ED
visitsh
owedreductions
from
27.5%
to36.7%
inch
ildren,
andbetw
een30.6%
to44%
foradults
None
None
Murphy
etal.
(1997)3
5
Before-and-after
study
Ireland
1993–1
995
Large,single
ED
Interventiongroup:Patients
were
GMS
ineligible
did
not
qualify
forfreeca
rein
Ireland
Controlgroup:allpatients
whovisitedED
Patients
whowere
GMS
ineligible
were
chargeda
certain
amountforeach
visit
totheED
insteadoftheir
GP.
Ifthepatienthadaletterof
referralfrom
aGP,they
were
exemptedfrom
thech
arge.
Thetotalnumberof
GMS
ineligible
visitedtheED
duringtheyear
before
theim
plantationofthe
copaymentwas19,562/43,202
(45.3%)andin
the
yearafterwas
19,947/45,302(44%),
whichsh
ows
areductionby1.3%
(95%
CI=–0
.6to
–1.9)
AmongtheGMS-ineligible
patients
referredbya
GPto
theED,thetotal
numberofthis
groupin
theyearbefore
was
2,619/19,562(13.4%)and
theyearafter3,156/19,947
(15.8%),anincrease
of
2.4%
(95%
CI=1.73to
3.1)
Notreported
Sch
illinger
etal.
(2000)3
8
Randomized
controltrial
UnitedStates
1997–1
998
Single,university-affiliated
primary
care
practice
Interventiongroup:
1,121patients
Controlgroup:1,172patients
Interventiongrouprequired
PCPapprovalforaccess
tosp
ecialtyandED
services
Nonsignifica
ntincrease
inrisk-adjusted,nonurgentED
visitsin
interventiongroup
(0.06visits/patient/year,
p=0.42)
PCPvisits:
nonsignifica
nt
increase
inintervention
group(0.27visits/patient/
year,
p=0.14)
Specialtyvisits:
significa
nt
reductionin
visitsin
interventiongroup
(0.57visits/patient/
year,
p=0.04)
Inpatientca
re:significa
ntreduction
inhosp
italizationsin
intervention
group(0.14hosp
italizations/
patient/year,
p=0.02)
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 977
Table
4Continued
Author(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:Capitation/PhysicianPayment
Catalano
etal.
(2000)2
8
Interrupted
time-series
quasi-
experiment
Pediatric
Medicaid
patients,
single
state
Interventiongroup:recipients
inareaswithca
pitation
Controlgroup:recipients
inareaswithoutca
pitation
Communitymentalhealth
centers
underca
pitated
agreementwithMedicaid.
Centers
billedbyhosp
ital
EDsforpsy
chiatric
emergency
services.
Nosignifica
ntincrease
inthelevelofED
usa
ge
Decrease
ininpatient
psy
chiatric
admission
Increase
inoutpatient
visitsin
forfor-profit
capitatedareas
Estim
ateddecrease
by$17–$
21
millionover1yearin
total
(inpatientandoutpatient)
costs
Catalano
etal.
(2005)2
9
Interrupted
time-series
quasi-
experiment
Single-state,Medicaid
patients
yearpriorandafter
intervention
Interventiongroup:recipients
inareaswithca
pitation
Controlgroup:recipients
inareaswithoutca
pitation
Communitymentalhealth
centers
underca
pitated
agreementwithMedicaid.
Centers
billedbyhosp
ital
EDsforpsy
chiatric
emergency
services
28%
declinein
psy
chiatric
ED
visitsco
mpared
toexpected
Notreported
Notreported
Dicke
yetal.
(1996)3
0
Before-
and-after
quasi-
experiment
Single-state,Medicaid
patients
Interventiongroup:
10,685patients
(1993)
7,541patients
(1994)
Controlgroup:
6,614patients
(1991)
7,295patients
(1992)
Single
providerofpsy
chiatric
services(inpatients
and
outpatient)
contractedwith
state
underca
pitatedagreement
19%
decrease
inED
visits
inyears
afterintervention
comparedto
years
prior
Nonquantifiedincrease
inoutpatientvisits
Inpatientadmission
decrease
dby4%
Medianlength-of-stay
decrease
dby3.3
days
Increase
in30-dayreadmissions
by1.9%
Reductionin
expenditures
perbeneficiary
$420
Glazier
etal.
(2009)3
2
Retrosp
ective
cohort
study
Canada
2005–2
006
Single
province
inco
untrywith
universalhealthca
resy
stem
Interventiongroup:487,131
patients
inprimary
care
practices
ofphysicianswhoch
ose
capitatedsy
stem
Controlgroup:
2,517,527patients
inprimary
care
practicesof
physicianswhoch
ose
fee-for-se
rvicesy
stem
Physiciansreim
bursedby
age-andse
x-adjusted
capitationpayments
and
requiredpatientenrollment
(stilleligible
forfee-for-se
rvice
payments
forhosp
italse
rvices)
Noch
angein
ED
visitsin
capitationgroupvs.
same
physicianspreca
pitation
Increase
like
lihoodof
OR
=1.20,95%
CI=1.15
to1.25)forca
pitation
groupvs.
fee-for-se
rvice
group(4%
more
patients
inca
pitationgroupmade
visitsto
ED
Higherratiooflow-acu
ity
visitsamongca
pitation
(1.6)vs.
control(0.9)groups
Less
after-hours
care
inca
pitationpractices
(OR
=0.68,95%
CI=0.61to
0.75)
Lowermorbidityandco
morbidity
indicesin
capitationpractices
Jose
phso
netal.
(1997)3
4
Retrosp
ective
cohort
study
Members
offiveinsu
rers
inonecity
Interventiongroups:
22,316
patients
infully
capitatedHMO
38,030patients
inany
ofthreeIPA
Controlgroup:14,110patients
inatraditionalindemnity
insu
rance
plan(estim
ated)
Capitationin
HMO.In
IPA,
capitatedpayments
tothe
associationbutfee-for-se
rvice
compensa
tionofphysician
ED
use
rates(p
<0.05):
70per1000members
inca
pitatedpanel,
363per1000members
inIPA
panel,
466per1000members
inindemnitypanel(85%
lowerED
use
ratesin
capitatedvs.
indemnitypanel)
None
Admissionforfiveambulatory
sensitiveco
nditions
0.8
per1000members
inca
pitatedpanel,
2.7
per1000members
inIPA
panel,
2.9
per1000members
inindemnitypanel
978 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
designed with physician or nurse gatekeepers who takeaway the patient’s autonomy by dictating whetherinsurance will cover an ED visit. Our findings, takenalong with prior reviews, are promising that non-EDinterventions designed to reduce ED visits may be suc-cessful; however, it is clear that more study is needed tounderstand the most effective ways to reduce ED use.
When organizations decide that reducing ED visits isa priority, the choice must be made which interventionsshould be implemented. We think that the choice shouldbe made based on organization priorities and consider-ing the profile of pros and cons for each intervention.For example, education is simple, can be inexpensive,and has an added benefit of improving health literacy;however, it is difficult to standardize. On the otherhand, managed care and patient financial incentives arepowerful tools but may have unintended consequences,like deferring needed care or limiting patient choice. Inaddition, adding capacity may have the opposite effectas desired. However, we can conclude that reducing EDuse will require broad, organizational changes. Changecould include a careful multilayered approach integrat-ing several interventions along with a feedback mecha-nism to monitor outcomes and adverse events.
Future research should attempt to delineate thebalance of intended and unintended effects of theseinterventions. Across all categories, more exploration ofthe effect on health outcomes and costs would be bene-ficial. Regarding education, researchers should considerfurther study of specific educational materials, frombooklet to in-person teaching, or even technology-basedsolutions. EMS diversion has such limited data that anydeeper examination of this area would add to ourknowledge, especially focusing on the safety and accu-racy of EMS assessments. Most of the managed carestudies are observations of large-scale systemic changesover several years. More randomized controlled trialsare needed across all categories to reduce confoundingand improve the generalizability of results.
LIMITATIONS
Many of the studies reported were observational withthe potential for confounders, and few studies reportedrisk-adjusted data. In addition, some of the paymentreform interventions were part of a bundle of changes,and it was impossible to unwind which specific inter-vention (or just the combination) was responsible forthe changes in utilization. Many studies were similarlyrestricted to a single site, which reduces the generaliz-ability to other settings.
Our primary audience was EDs in English-speakingparts of the world; therefore, we included only studiespublished in English. As a result, this study may missimportant findings from non-English studies. Also, eightof the 39 studies were based in Canada or Europe,while the rest were in the United States, and differencesin health care systems, in particular the presence ofsingle payer versus multipayer systems, may affectoutcomes.
There was also variability in the interventions admin-istered as well as the outcomes reported. For example,some studies reported total ED visits whereas othersT
able
4Continued
Author
(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Piehl
etal.
(2000)3
6
Before-and-after
study
UnitedStates
1995–1
997
Pediatric
visitsattw
oco
mmunity
EDsover2year,
determ
inedrates
ofED
visitsafterim
plementation
ofmanagedca
replan
Interventiongroup:
Medicaid-m
anagedca
rewith20,663
ED
visitsControlgroup:non-M
edicaid
group:34,079ED
visits
Managedca
reprogram
(recipients
haveassigned
PCPand24-houraccess,at
least
telephone);PCPrece
ived
monthly
feeforeach
enrollee
inadditionto
fee
foreach
service
Medicaid
group:24%
reductionin
overall
ED
visits(p
<0.001)(33.5
[SD
�5.3]
reduce
dto
25.6
[SD
�2.3]monthly
ED
visitper1,000enrollees);37%
reduction
innonurgentED
visits(p
<0.001)
Non–M
edicaid-insu
redgroup:
8%
increase
inoverallED
visits(p
<0.001)
8%
increase
innonurgentED
visits(p
notgiven)
Notreported
Notreported
Price etal.
(1999)3
7
Retrosp
ective
cohort
study
UnitedStates
Oneyear
Threeasthmatic
pediatric
patients
groupsbase
dontheir
typeofinsu
rance
Fee-for-se
rvicegroup:
47patients
Capitated(H
MO)group:
24patients
Medicaid
group:22patients
Theeffect
ofthetypeof
insu
rance
onthemedical
care
use
in1year
ED
use
:Feeforse
rvice,2/47(4%);
Capitated,6/24(25%);
Medicaid,7.5/22(34%);
p=0.038
Physicianvisits
Feeforse
rvice,12/47
(25%);Capitated,12/24
(50%);Medicaid,9.5/22
(43%);p=0.56
Specialist
visits
Feeforse
rvice,6/47
(12.7%);Capitated,
7.5/24(31%);Medicaid,
3/22(13.6%);p=0.118
Hosp
italvisits
Feeforse
rvice,2/47(4%);
Capitated,2/24(8.3%);
Medicaid,3/22(13.6%);
p=0.14
AFDC
=Aid
toFamilieswithDependentChildren;GMS
=GeneralMedicalServices;
GP=generalpractitioner;
HMO
=healthmaintenance
organization;IPA
=independentprac-
tice
associations;
PCP=primary
care
clinic.
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 979
Table
5PatientFinancialInce
ntives
Author(Year)
Design
Population
Cost
Intervention
Effect
onED
Use
Effect
on
Non-ED
Use
OtherOutcomes
Intervention:Copayment/Coinsu
rance
Hsu
etal.
(2006)3
9Quasi-experimental,
longitudinal,
concu
rrentco
ntrols
UnitedStates
1999–2
001
Riskadjusted
Prepaid
integrateddelivery
system,
19medicalce
nter
Interventiongroup:
commercially
insu
red,2,257,445patients
(copaymentlevels:$0,$1–$
5,
$10–$
15,
$20–$
35,and$50–$
10perED
visit)
Controlgroup:
Medicare,withemployer
supplementation,261,091
patients
(copaymentlevels:$0,$1–$
15,
and$20–$
50perED
visit)
•Compariso
nofvarious
copayment
levels
over36-m
onth
timeperiod
•Copaymentlevelch
ose
nbyemployer,
notpatient
•In
commerciallyinsu
red
group,risk-adjustedRR
of
ED
visitco
mparedto
no
copaymentgroupwas0.96
for$1–5
copaymentgroup
(CI=0.96–0
.97),0.93for$10
–15(CI=0.92–0
.94)0.88for
$20–3
5(CI=0.87–0
.89)0.77
for$50–1
00(CI=0.76–0
.77)
•In
Medicare
group,RR
ofED
visitco
mparedto
noco
pay-
mentgroupwas0.97for$1–
15co
paymentgroup
(CI=0.96–9
9)0.96for$20–
50(CI=0.94–0
.97)
Notreported
•Nosignifica
ntch
angein
hosp
italizations,
ICU
admission,anddeathswith
higherco
paymentrates
exce
ptthefollowing:
•In
commerciallyinsu
red
group,death
rate
was
significa
ntlyhigherin
lowest
copaymentlevelof
$1–$
5co
mparedto
no
copaymentwithRR
=1.09
(CI=1.02–1
.16)
•In
Medicare
group,highest
copaylevel($20–$
50)with
decrease
dmortality
rate
(RR
=0.87,CI=0.83–0
.91)
Loweetal.
(2008)4
7Before-and-after
obse
rvational
study
UnitedStates
48months
26EDs,
purposivesa
mple
of
thestate
Interventiongroup:post–
Oregon
healthplancu
tbacks
Controlgroup:pre–O
regon
healthplancu
t-backs
Fivegroupsofpatients
base
dontheclass
ofpayer:
•Commerciallyinsu
red
•Oregonhealthplan
•Uninsu
red
•Medicare
•Others
payers
Medicaid
cutback,Oregon:the
beneficiariesoftheOregon
healthplanrece
ivenew
policy
changesthatinclude
copaymentformost
ofthe
healthse
rvices(PCPvisits,
ED
use
,andhosp
italization);
decrease
denrollmentin
healthplanwithincrease
dnumberofuninsu
redpatients
•Oregon
health
plan
20%
reductionin
ED
visits/month
(CI=13–1
8)
•Uninsu
red
20%
increase
inED
visits/
month
(CI=13to
18)
Notreported
Hosp
italizationfrom
ED
visit
(adjustedORs):
•Commercial
0.99(CI=0.95–1
.04)
•Oregonhealthplan
1.09(CI=1.03–1
.16)
•Uninsu
red
1.50(CI=1.39–1
.62)
•Medicare
1.10(CI=1.06–1
.13)
Loweetal.
(2010)4
0Before-and-after
study
UnitedStates
2001–2
004
State
populationstudy
Interventiongroup:
Medicaid
enrolleeswiththe
standard
plan
Controlgroup:Medicaid
enrollees
withtheplusplan(not
affected
bycu
tbacks)
Medicaid
cutback,Oregon:
Thebeneficiariesofthe
Oregonhealthplanrece
ive
new
policy
changesthat
includeco
payment
formost
ofthehealth
services(PCPvisits,
ED
use
,andhosp
italizationand
eliminatedoutpatient
behavioralhealth
services
Decrease
inED
use
rates
(RR
=0.84;95%
CI=0.83–
0.86).Comparedto
PlusPlan
members,use
alsodecrease
d(RR
=0.82;95%
CI=0.80–
0.84)
Injury-relatedvisitsalso
decrease
d(p
<0.001).
Notreported
ED
visitsleadingto
hosp
ital
admissionalsodecrease
d:
(RR
=0.83;95%
CI=0.79–
0.86).Comparedto
Plus
Planmembers,utilization
alsodecrease
d(RR
=0.85;
95%
CI=0.82–0
.89).
Mortense
n(2010)4
8Quasi
experimental,
pre–p
ost
design
(difference
s-in-
difference
smethodology)
24months
Riskadjusted
State-levelstudyamong
Medicaid
enrolleesin
29states.
Interventiongroup:Medicaid
enrollees
instateswithco
payment
increase
(ninestates)
Controlgroup:Medicaid
enrolleesin
stateswithoutno
copaymentornoco
payment
change(20states)
•Medicaid
ED
visit
copayment
increase
inninestates,
rangingfrom
$3to
$50
Significa
ntincrease
inprobabilityofanyED
visitsin
ayear(33.1
perperson-m
onth
forch
angegroupvs.
24.7
for
controlgroup,p=0.000)
Notreported
Notreported
980 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
Table
5Continued
Author(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:HDHP
O’G
rady
etal.
(1985)4
1
Randomized
controlledtrial
UnitedStates
State-levelamongsix
geographic
areasin
fourstates
Interventiongroup:
3,973personswere
assignedrandomly
todifferentfee-for-
serviceinsu
rance
planswithco
payment
ratesof0,25,50,or95%
Aco
insu
rance
rate
of
25,50,or95%—
upto
maxim
um
of$1,000
asout-of-pocket
costs
TheprobabilityofED
use
amongtheco
payment
groupswasfewerthanthefreegroup
•25%
copaymentgroupfewerby15%
(p=0.05)
•50%
copaymentgroupfewerby8%
•95%
copaymentgroupfewerby30%
(p=0.01)
•Individual-deductible
planfewerby19%
(p=0.05)
•TheED
use
amongtheco
paymentgroupswas
fewerthanthefreegroup
•25%
copaymentgroupfewerby21%
(p=0.01)
•50%
copaymentgroupfewerby18%
•95%
copaymentgroupfewerby35%
(p=0.01)
•Individual-deductible
planfewerby20%
(p=0.05)
Notreported
TheED
visitresu
ltingin
hosp
italizationamong
theco
paymentgroups
wasfewerthanthefree
group
•25,50,and95%
copaymentgroups
fewerby33%
(p<0.05)
Selbyetal.
(1996)4
2Matchedco
hort
study
Riskadjusted
Group-m
odelHMO,
15medicalce
nters
Interventiongroup:
30,276patients
Controlgroup:60,408
patients
(matchedfor
age,se
x,andloca
tion);
seco
ndgroup37,539
patients
(matched
forsa
me+employer)
Introductionofa
$25–$
35
copaymentforED
use
inanHMO
•Abso
lute
decrease
inED
usa
geby28visits/
100person-yearin
copaymentgroup
•Risk-adjusteddecrease
inED
visits14.6%
more
inco
paymentgroupthaneitherco
ntrol
group(p
<0.001)
•Significa
ntdecrease
sin
visitsforless
emer
gentco
nditionsco
mparedto
emergent
•Nosignifica
ntch
angein
urgentca
reuse
andpediatric
office
visitsovertimebetw
een
groups
•4.4%
decrease
inrate
of
adultoffice
visitsin
copayment
group
•Nonsignifica
ntreduction
inpotentiallyavoidable
hosp
italizationsin
copay
mentgroupco
mparedto
controlgroups
•Significa
ntlylower
adjustedmortality
rate
intheco
paymentgroup
(1.6/1,000)thanin
control
group1(2.2/1,000,
p=0.001)andco
ntrol
group2(2.6/1,000,
p=0.06)
Wallace
etal.
(2008)4
3Retrosp
ective
cohort
study,
difference
-in-
difference
smethodology
UnitedStates
2001–2
004
State
populationstudy,
propensity
score
matchedpopulation
Interventiongroup:
Medicaid
enrollees
withtheStandard
Plan,10,176su
bjects
ControlGroup:
Medicaid
enrollees
withthePlusPlan,
10,319su
bjects
Additionof$50
copaymentforED
visitsthatdid
not
resu
ltin
admission.
Interventionalso
includedelimination
ofce
rtain
benefits
(dental,mentalhealth,
etc.)aswellasother
copayments
(outpatientvisits,
inpatientadmissions,
pharm
acy
,etc).
ED
visitsdecrease
dby7.9%
(p=0.03)in
the
standard
(copayment)
grouprelativeto
the
plusgroup.ED
visitexpendituresbyuse
rincrease
d7.9%
(p=0.03)overthesa
me
timeperioddesp
itethedecrease
Invisits.
Pharm
acy
:decrease
inuse
and
expenditures(–2.2%,
p<0.001;–1
0.5%,p<0.001).
Inpatient:increase
inuse
and
expenditures(+27.3%,p<0.001;
+20.1,p<0.001)
Hosp
italoutpatient:
increase
inuse
andexpenditures
(+13.5%,p<0.001;19.7%,
p<0.001)
Ambulatory:decrease
inuse
(–7.7%,p<0.001)butincrease
inexpendituresbyuse
r(+6.6%,
p=0.75)
Totalexpenditures(+2.2%,
p=0.47)byperson
unch
angeddesp
iteoverall
reductionin
use
(–2.2%,
p<0.001)
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 981
Table
5Continued
Author(Year)
Design
Population
Intervention
Effect
onED
Use
Effect
onNon-ED
Use
OtherOutcomes
Intervention:HDHP
Waters
etal.
(2011)45
Retrosp
ective
cohort
study
Riskadjusted
Majorinsu
rerin
single
state,propensity
score
matchedsa
mple
Interventiongroup:1,354HDHPgroup
initiallyin
PPO,thensw
itch
ed
andmaintained
inHDHPforall3studyyears
Controlgroup:1,354members
of
enrolledin
PPO
plan
HDHPwithdeductiblesranging
from
$1,700to
$6,000vs.
PPO
plan.HDHPenrolleesuse
dstandard
PPO
contracted
amounts
before
meeting
their
deductible.
HDHPenrolleeshad
lowerprobabilityof
use
andlevelofuse
(p<0.05)
HDHPenrolleesappeared
tohavedecrease
dprimary
care
use
(p<0.01),but
greatersp
ecialtyphysician
use
(p<0.05)
HDHPenrolleeshadsignifica
ntly
higherprescriptiondrug
utilizationandexpenditures
(p<0.05)
Wharam
etal.
(2007)44
Before-and-after
study
UnitedStates
2001–2
005
Single-state,single-insu
rance
carrier
offeringHDHPandHMO
plans
Interventiongroup:8,724enrolleeswith
HDHPforatleast
6months,
after1year
oftraditionalHMO
Controlgroup:59,557enrolleeswith
traditionalHMO,matchedon
adult/childstatus.
Individuals
intheHDHPgroup
haveannualindividual
deductiblesrangingfrom
$500to
$2,000.Ifthe
expenditure
exce
edsthedeductible,
individuals
payaco
payment
($100)foreach
ED
use
.Individuals
intheintervention
grouphadco
paymentfor
each
ED
($20-$100)and
outpatientvisit($5–$
25).
ED
visitswith10.0%
relativedecrease
(abso
lute
change,20.2
visitsper1,000)in
the
HDHPgroupco
mpared
withco
ntrols
from
base
lineto
follow-up
(95%
CI=–1
6.6%
to–2
.8%;p=0.007)
ED
visitsresu
ltingin
hosp
italizationswith24.7%
relativedecrease
(withan
abso
lute
rate
difference
of
�2.6%
intheHDHPgroup
(95%
CI=�4
1.0%
to�3
.9%;
p=0.02)
ED
expense
perHDHPmember
decrease
dfrom
$75in
the
base
lineyearto
$36in
the
follow-upperiod,anabso
lute
declineof$52,represe
ntinga
58.5%
relativedecline(95%
CI=–6
4.4%
to–5
1.5%;
p<0.001)
Intervention:HealthSavingsAccount
Wilso
netal.
(2008)46
Multiyearcross-
sectionalstudy
UnitedStates
2004–2
006
Riskadjusted
Commercialinsu
rance
planwithboth
comprehensivemajor
medicalandCDHP
Interventiongroup:enrolleesofCDHP
planforthestudyyear
Controlgroup:enrolleesoftraditional
comprehensivemajormedicalplan
CDHPincludesavariety
of
productsincludinghealth
reim
bursementaccounts
and
healthsa
vingsaccounts,
whichare
portable
ifan
employeeleaves
his
orheremployer.
Alloffer
first-dollarco
veragefor
preventivese
rvices.
Afterrisk
adjustment,
CDHPplanmembers
had
129.1
vs.
141.2
ED
visits/
1,000members/year
CDHP=co
nsu
mer-driven
health
plan;HDHP=high-deductible
health
plan;HMO
=health
maintenance
organization;ICU
=intesive
care
unit;PCP=primary
care
provider;
PPO
=preferredproviderorganization;RR
=rate
ratio.
982 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
reported ED visits per user. This made comparisonbetween the studies difficult and limits the ability todraw robust conclusions from the findings of thereview.
CONCLUSIONS
We found that about two-thirds of the studied interven-tions showed reductions in ED use. Future studiesshould attempt to reduce confounding through robustdesign (i.e., randomization) and should measure unin-tended consequences of increased demand for othertypes of health care services, health outcomes, andfinancial effects.
References
1. Centers for Medicare & Medicaid Services. NationalHealth Expenditure Tables. Available at: http://www.cms.gov/Research-Statistics-Data-and-Systems/Stati-stics-Trends-and-Reports/NationalHealthExpendData/downloads/tables.pdf. Accessed Jul 25, 2013.
2. Asplin BR, Magid DJ, Rhodes KV, Solberg LI, LurieN, Camargo CA Jr. A conceptual model of emer-gency department crowding. Ann Emerg Med.2003; 42:173–80.
3. Mehrotra A, Liu H, Adams JL, et al. Comparingcosts and quality of care at retail clinics with that ofother medical settings for 3 common illnesses. AnnIntern Med. 2009; 151:321–8.
4. Katz EB, Carrier ER, Umscheid CA, Pines JM. Com-parative effectiveness of care coordination interven-tions in the emergency department: a systematicreview. Ann Emerg Med. 2012; 60:12–23.
5. Bunn F, Byrne G, Kendall S. Telephone consultationand triage: effects on health care use and patientsatisfaction. Cochrane Database Syst Rev. 2004; (4):CD004180.
6. Althaus F, Paroz S, Hugli O, et al. Effectiveness ofinterventions targeting frequent users of emergencydepartments: a systematic review. Ann Emerg Med.2011; 58:41–52.
7. Flores-Mateo G, Violan-Fors C, Carrillo-SantisteveP, Peir�o S, Argimon JM. Effectiveness of organiza-tional interventions to reduce emergency depart-ment utilization: a systematic review. PLoS ONE.2012; 7:e35903.
8. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMAstatement for reporting systematic reviews andmeta-analyses of studies that evaluate health careinterventions: explanation and elaboration. BMJ.2009; 339:b2700.
9. Guyatt GH, Oxman AD, Vist G, et al. GRADE guide-lines: 4. Rating the quality of evidence–study limita-tions (risk of bias). J Clin Epidemiol. 2001; 64:407–15.
10. Glavan KA, Haynes M, Jones DR, Philput C. A mili-tary application of a medical self-care program. MilMed. 1998; 163:678–81.
11. McWilliams DB, Jacobson RM, Van Houten HK,Naessens JM, Ytterberg KL. A program of anticipa-tory guidance for the prevention of emergencydepartment visits for ear pain. Arch Pediatr AdolescMed. 2008; 162:151–6.
12. Rettig BA, Shrauger DG, Recker RR, Gallagher TF,Wiltse H. A randomized study of the effects of ahome diabetes education program. Diabetes Care.1986; 9:173–8.
13. Schonlau M, Mangione-Smith R, Chan KS, et al.Evaluation of a quality improvement collaborative inasthma care: does it improve processes and out-comes of care? Ann Fam Med. 2005; 3:200–8.
14. Rector TS, Venus PJ, Laine AJ. Impact of mailinginformation about nonurgent care on emergencydepartment visits by Medicaid beneficiaries enrolledin managed care. Am J Manag Care. 1999; 5:1505–12.
15. Hudec JC, MacDougall S, Rankin E. Advancedaccess appointments: effects on family physician
Table 6Quality Assessment of Selected Observational Studies on Non-ED Interventions to Reduce ED Utilization Using the Grading ofRecommendations Assessment, Development, and Evaluation(GRADE) Criteria
Possible limits of observational studies
1. Failure to develop and apply appropriate eligibility criteria(inclusion of control population)
2. Flawed measurement of both exposure and outcome3. Failure to control confounders and to measure all known
prognostic factors4. Imprecision of outcomes (i.e., wide CIs)5. Incomplete follow-up
First Author (Reference)Existing
LimitationsOverall Qualityof Evidence
Patient educationGlavan (10) 1, 3, 4 Very lowMcWilliams (11) 1, 3 Very lowSchonlau (13) 1, 3 Very low
Non-ED capacity increaseChalder (22) 1, 2, 3 Very lowHsu (24) 1, 3, 4 Very lowHudec (15) 1, 3, 4 Very lowO’Kelly (16) 1, 2, 3, 4 Very lowPhilips (23) 1, 2, 3 Very lowRust (19) 1, 2, 3 Very lowSolberg (20) 1,2,3 Very lowSolberg (21) 1,2,3,4 Very lowvan Uden (17) 1, 2, 3 Very lowWang (18) 2, 3 Very low
Prehospital diversionSchaefer (25) 1, 3, 5 Very low
Managed careBadgett (27) 1, 3, 4 Very lowCatalano (28) 1, 3 Very lowCatalano (29) 1,3, 4 Very lowDickey (30) 1, 3, 4 Very lowFranco (31) 1, 2, 3 Very lowGlazier (32) 1, 3 Very lowHurley (33) 1, 2, 3, 4 Very lowJosephson (34) 1,3, 4 Very lowMurphy (35) 1, 3 Very lowPiehl (36) 3, 4 Very lowPrice (37) 2, 4 Very low
Patient financial incentivesHsu (39) 3 Very lowLowe (47) 1, 2, 3 Very lowLowe (40) 2, 3 Very lowMortensen (48) 1, 2, 3, 5 Very lowSelby (42) 3 Very lowWallace (43) 3 Very lowWaters (45) 2, 3,4 Very lowWharam (44) 1,3,5 Very lowWilson (46) 2, 3 Very low
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 983
satisfaction, physicians’ office income, and emer-gency department use. Can Fam Phys. 2010; 56:e361–7.
16. O’Kelly FD, Teljeur C, Carter I, Plunkett PK. Impactof a GP cooperative on lower acuity emergencydepartment attendances. Emerg Med J. 2010;27:770–3.
17. van Uden CJ, Crebolder HF. Does setting up outof hours primary care cooperatives outside a hospi-tal reduce demand for emergency care? Emerg MedJ. 2004; 21:722–3.
18. Wang C, Villar ME, Mulligan DA, Hansen T. Costand utilization analysis of a pediatric emergencydepartment diversion project. Pediatrics. 2005;116:1075–9.
19. Rust G, Baltrus P, Ye J, et al. Presence of a commu-nity health center and uninsured emergency depart-ment visit rates in rural counties. J Rural Health.2009; 25:8–16.
20. Solberg LI, Maciosek MV, Sperl-Hillen JM, et al.Does improved access to care affect utilization andcosts for patients with chronic conditions? Am JManag Care. 2004; 10:717–22.
21. Solberg LI, Crain AL, Sperl-Hillen JM, HroscikoskiMC, Engebretson KI, O’Connor PJ. Effect ofimproved primary care access on quality of depres-sion care. Ann Fam Med. 2006; 4:69–74.
22. Chalder M, Sharp D, Moore L, Salisbury C. Impactof NHS walk-in centres on the workload of otherlocal healthcare providers: time series analysis. BrMed J. 2003; 326:532–4.
23. Philips H, Remmen R, Van Royen P, et al. What’sthe effect of the implementation of general practi-tioner cooperatives on caseload? Prospective inter-vention study on primary and secondary care. BMCHealth Serv Res. 2010; 10:222.
24. Hsu RT, Lambert PC, Dixon-Woods M, KurinczukJJ. Effect of NHS walk-in centre on local primaryhealthcare services: before and after observationalstudy. Br Med J. 2003; 326:530–2.
25. Schaefer RA, Rea TD, Plorde M, Peiguss K, Gold-berg P, Murray JA. An emergency medical servicesprogram of alternate destination of patient care.Prehosp Emerg Care. 2002; 6:309–14.
26. Snooks H, Foster T, Nicholl J. Results of an evalua-tion of the effectiveness of triage and direct trans-portation to minor injuries units by ambulancecrews. Emerg Med J. 2004; 21:105–11.
27. Badgett JT. Can Medicaid format alter emergencydepartment utilization patterns? Pediatr EmergCare. 1986; 2:67–70.
28. Catalano R, Libby A, Snowden L, Cuellar AE. Theeffect of capitated financing on mental health ser-vices for children and youth: the Colorado experi-ence. Am J Public Health. 2000; 90:1861–5.
29. Catalano RA, Coffman JM, Bloom JR, Ma Y, KangSH. The impact of capitated financing on psychiatricemergency services. Psychiatric Serv. 2005; 56:685–90.
30. Dickey B, Normand S-T, Norton EC, Azeni H,Fisher W, Altaffer F. Managing the care of schizo-phrenia: lessons from a 4-year Massachusetts Med-icaid study. Arch Gen Psychiatry. 1996; 53:945–52.
31. Franco SM, Mitchell CK, Buzon RM. Primary carephysician access and gatekeeping: a key to reducingemergency department use. Clin Pediatr. 1997;36:63–8.
32. Glazier RH, Klein-Geltink J, Kopp A, Sibley LM.Capitation and enhanced fee-for-service models forprimary care reform: a population-based evaluation.CMAJ. 2009; 180:E72–81.
33. Hurley RE, Freund DA, Taylor DE. Emergencyroom use and primary care case management: evi-dence from four Medicaid demonstration programs.Am J Public Health. 1989; 79:843–6.
34. Josephson GW, Karcz A. The impact of physicianeconomic incentives on admission rates of patientswith ambulatory sensitive conditions: an analysiscomparing two managed care structures andindemnity insurance. Am J Manag Care. 1997; 3:49–56.
35. Murphy AW, Leonard C, Plunkett PK, et al. Effect ofthe introduction of a financial incentive for fee-pay-ing A and E attenders to consult their general prac-titioner before attending the A and E department.Fam Pract. 1997; 14:407–10.
36. Piehl MD, Clemens CJ, Joines JD. Narrowing thegap: decreasing emergency department use by chil-dren enrolled in the Medicaid program by improv-ing access to primary care. Arch Pediatr AdolescMed. 2000; 154:791–5.
37. Price MR, Norris JM, Bartleson BB, Gavin LA,Klinnert MD. An investigation of the medical careutilization of children with severe asthma accordingto their type of insurance. J Asthma. 1999; 36:271–9.
38. Schillinger D, Bibbins-Domingo K, Vranizan K, Bac-chetti P, Luce JM, Bindman AB. Effects of primarycare coordination on public hospital patients. J GenIntern Med. 2000; 15:329–36.
39. Hsu J, Price M, Brand R, et al. Cost-sharing foremergency care and unfavorable clinical events:findings from the safety and financial ramificationsof ED copayments study. Health Serv Res. 2006;41:1801–20.
40. Lowe RA, Fu R, Gallia CA. Impact of policy changeson emergency department use by Medicaid enrol-lees in Oregon. Med Care. 2010; 48:619–27.
41. O’Grady KF, Manning WG, Newhouse JP, BrookRH. The impact of cost sharing on emergencydepartment use. N Engl J Med. 1985; 313:484–90.
42. Selby JV, Fireman BH, Swain BE. Effect of a copay-ment on use of the emergency department in ahealth maintenance organization. N Engl J Med.1996; 334:635–41.
43. Wallace NT, McConnell KJ, Gallia CA, Smith JA.How effective are copayments in reducing expendi-tures for low-income adult Medicaid beneficiaries?Experience from the Oregon health plan. HealthServ Res. 2008; 43:515–30.
44. Wharam JF, Landon BE, Galbraith AA, KleinmanKP, Soumerai SB, Ross-Degnan D. Emergencydepartment use and subsequent hospitalizationsamong members of a high-deductible health plan. JAm Med Assoc. 2007; 297:1093–102.
45. Waters TM, Chang CF, Cecil WT, Kasteridis P, Mir-vis D. Impact of high-deductible health plans on
984 Morgan et al. • SYSTEMATIC REVIEW OF NON-ED INTERVENTIONS TO REDUCE ED USE
health care utilization and costs. Health Serv Res.2011; 46:155–72.
46. Wilson AR, Bargman EP, Pederson D, et al. Morepreventive care, and fewer emergency room visitsand prescription drugs–health care utilization in aconsumer-driven health plan. Benefits Q. 2008;24:46–54.
47. Lowe RA, McConnell KJ, Vogt ME, Smith JA.Impact of Medicaid cutbacks on emergency depart-ment use: the Oregon experience. Ann Emerg Med.2008; 52:626–34.
48. Mortensen K. Copayments did not reduce Medicaidenrollees’ nonemergency use of emergency depart-ments. Health Aff (Millwood). 2010; 29:1643–50.
APPENDIX
MEDLINE SEARCH TERMS
(((“emergency medical services” [all fields] OR “emer-gency medicine” [all fields] OR “emergency department”[all fields]) AND ((“utilization” [subheading] OR “utiliza-
tion” [all fields]) OR demand [all fields] OR visits [allfields])) AND (“intervention” [all fields] OR “solutions”[MeSH Terms] OR “solutions” [all fields])) AND(“advanced access” [all fields] OR “primary care” [allfields] OR capacity [all fields] OR “extended hours” [allfields] OR “telephone triage” [all fields] OR “care man-agement” [all fields] OR “patient education” [all fields]OR “general practitioner” [all fields] OR “care coordina-tion” [all fields] OR copayment [all fields] OR diversion[all fields] OR acuity [all fields]) AND English [lang]AND English [lang].
Supporting Information
The following supporting information is available in theonline version of this paper:
Data Supplement S1. PRIMSA flow diagram.Data Supplement S2. Quality assessment of selected
randomized trials on non-ED interventions to reduceED use using the Grading of Recommendations Assess-ment, Development, and Evaluation (GRADE) criteria.
ACADEMIC EMERGENCY MEDICINE • October 2013, Vol. 20, No. 10 • www.aemj.org 985