Outcomes of Community-dwelling Seniors Vary by Type of Emergency Department

ORIGINAL RESEARCH CONTRIBUTION

Outcomes of Community-dwelling SeniorsVary by Type of Emergency DepartmentJane McCusker, MD, DrPH, Danièle Roberge, PhD, Antonio Ciampi, PhD,Roxane Borges Da Silva, PhD, Alain Vadeboncoeur, MD, Danielle Larouche, MSc,Jean-Frédéric Lévesque, MD, PhD, and Eric Belzile, MSc

AbstractObjectives: The specific objectives were: 1) to compare the characteristics and 6-month outcomes ofcommunity-dwelling seniors in Quebec, Canada, who visited three different emergency department (ED)types and 2) to explore whether the differences in outcomes by ED type were seen among subgroups ofseniors.

Methods: The three types of ED were most specialized, less community-oriented (n = 12); moderatelyspecialized, less community-oriented (n = 28); and least specialized, more community-oriented (n = 28).Administrative databases were used to create a cohort of 223,120 seniors who visited these 68 EDs dur-ing a 14-month period. Using a multilevel approach, the following patient characteristics were comparedacross ED types: sociodemographic (age, sex, urban vs. rural residence, proximity to ED); medical diag-noses and comorbidity burden; and utilization of hospital and physician services during the 16 monthsbefore the index ED visit. Cox regression analysis was used to model the relationships between ED typeand two 6-month outcomes, adjusting for patient characteristics: 1) serious outcomes (death, acute orlong-term-care admission) among all individuals who made an index visit and 2) outpatient ED visits(without hospital admission) among those discharged either from the ED or hospital. Interactionsbetween ED type and patient age, sex, urban–rural residence, and comorbidity burden were explored.

Results: Compared to patients treated at the least specialized EDs, those at the most specialized EDswere more often urban-dwelling, resided outside the health service area of the ED, and had the highestdisease burden and prior specialist utilization. Those treated at the moderately specialized EDs wereintermediate between these two groups. During the 6 months after the ED visit, the rate of serious out-comes was higher and the rate of outpatient ED visits was lower for the most specialized compared tothe least specialized EDs, even after adjustment for patient characteristics. The differences in theseoutcomes by ED type were attenuated among older patients and those with greater comorbidity.

Conclusions: More vulnerable community-dwelling seniors tend to be treated in more specialized EDs,which have worse linkages to community services. Improved linkages between more specialized EDsand the community (physicians, home care, and other services) and increased access to community ser-vices may improve outcomes in this population. Seniors treated at more specialized EDs were morelikely to experience serious outcomes, but were less likely to make a return outpatient ED visit.

ACADEMIC EMERGENCY MEDICINE 2012; 19:304–312 ª 2012 by the Society for Academic EmergencyMedicine

ISSN 1069-6563 ª 2012 by the Society for Academic Emergency Medicine304 PII ISSN 1069-6563583 doi: 10.1111/j.1553-2712.2012.01295.x

From the Department of Epidemiology, Biostatistics and Occupational Health (JM, AC), McGill University, and St Mary’sResearch Centre (JM, AC, EB), Montreal, Quebec; the Centre de Recherche de l’Hôpital Charles LeMoyne (DR, DL), Longueuil,Quebec; the Université de Sherbrooke (DR), Sherbrooke, Quebec; McGill University (RBDS), Montreal, Quebec; Emergency Medi-cine Services, Montreal Institute of Cardiology (AV), Montreal, Quebec; and the Centre de recherche du CHUM et InstitutNational de Santé Publique du Québec (JFL), Montréal, Quebec, Canada.Received June 1, 2011; revision received August 30, 2011; accepted September 1, 2011.This work was presented in part at the Canadian Association for Health Services and Policy Research (CAHSPR) Annual Confer-ence, Toronto, Ontario, May 2010.This work was financed by a grant from Fonds de la Recherche en Santé du Québec. RDS was funded by the Canadian Institutes ofHealth Research and Réseau de Recherche en Santé des Populations du Québec Strategic Training Program in TransdisciplinaryResearch on Public and Population Health Interventions: Promotion, Prevention and Public Policy (4P).The authors have no relevant financial information or potential conflicts of interest to disclose.Supervising Editor: Manish Shah, MD, MPH.Address for correspondence and reprints: Jane McCusker, MD, DrPH; e-mail: [email protected].

S eniors (defined here as aged 65 years or over) arean important population served by emergencydepartments (EDs), because of both the greater

complexity of their care and their worse outcomes incomparison with younger adults.1 The care they receiveat and after an ED visit (e.g., ED-based screening andassessment of needs, plus referral to appropriate com-munity-based services) can affect their outcomes afterthe visit.2–6 It is therefore important to determinewhether more vulnerable seniors receive appropriategeriatric services at an ED visit. Because EDs varygreatly in their resources and organization, population-based studies are needed to address this question.

Most prior research on ED care and outcomesamong seniors has been conducted in a small numberof EDs, typically large, urban, and university-affili-ated.1,5 The results of two recent population-basedstudies suggest that there are differences in outcomesof seniors by ED size, and provision of certain geriatricED services.7,8 This research has not addressed com-prehensively the relationship between the characteris-tics of seniors making an ED visit and the geriatricservices provided by the ED. Therefore, in this studywe used an empirically derived classification of EDsinto three broad types, based on their staffing, organi-zation of ED geriatric services, and linkages to commu-nity resources for the care of seniors.9 The three EDtypes differ in their internal specialization of staffingand care processes and their external orientation (link-ages to community physicians and community-basedgeriatric services). Type 1 EDs are the most internallyspecialized but with weak links to community resources.Type 3 EDs are the least internally specialized, but arebetter linked to community resources. Type 2 EDs areintermediate in their characteristics and generally moreheterogeneous than the other types.

In this study, we aimed to describe and compare thecharacteristics and 6-month outcomes of community-dwelling seniors who visited these three types of ED. Inthis observational study, based on administrative data-bases, we were interested in certain outcomes (survival,hospital admission, or long-term care [LTC] institutional-ization) as indicators of vulnerability, rather than as out-comes that might be affected by the type of ED visited.We were interested in particular in whether the mostvulnerable seniors were more likely to visit a particulartype of ED and whether their outcomes differed by EDtype. We therefore examined four subgroups of seniors:the oldest old (aged 85+ years), women, those living inurban areas (an indicator of relatively poor access tofamily doctors),10 and those with a higher comorbidityburden. Finally, we aimed to describe return outpatientED visits in this population, as an indicator of use of theED as an alternative source of primary care.11

METHODS

Study DesignWe used Canadian provincial administrative databasesto define a cohort of community-dwelling seniors whovisited the 68 EDs during a 14-month period. Patientcharacteristics were measured during the 16 monthsbefore the index ED visit for each patient, and outcomes

were measured during the 6 months after the index EDvisit. The study protocol was approved by the provin-cial body responsible for approving the use of adminis-trative databases for research (Commission d’Accès àl’Information) and the McGill University Research Eth-ics Committee.

Study Setting and PopulationThis research was conducted in the province of Que-bec, Canada, a population with universal health insur-ance. International comparative studies have found thatthe Quebec population has among the highest rates ofED visits and some of the longest waiting times toreceive care in an ED.12–14

The patient cohort was created from three individuallylinked administrative databases for the governmentalhealth insurance program (enrollee file, hospital dis-charges, and physician billings) for the Quebec popula-tion for a 3-year period: April 1, 2003, to March 31, 2006.The derivation of the study sample is shown in Figure 1.Briefly, the sample consisted of all patients aged 65 yearsand over with an index ED visit at one of the 68 EDs dur-ing the 14-month period August 1, 2004, to September30, 2005. Because Quebec administrative databases donot allow linkage of ED visit data from registries withdata on patient characteristics and outcomes, we definedan ED visit as one or more ED billings on up to 2consecutive days; this definition gives the best overallagreement with ED-level visits from registries.15

We defined an index visit as one with no ED billingsin the previous 30 days, to reduce the likelihood thatthe index visit was a follow-up to a previous visit. Sub-jects were excluded if they lived in a LTC facility beforethe index visit or visited a psychiatric hospital ED or ifwe were unable to identify the ED from which theywere discharged. We excluded 35 EDs with insufficientdata to be classified; these EDs were similar to the 68EDs in the sample with respect to their organizationaland other characteristics.9 The main cohort for theanalysis of serious outcomes included 223,120 patientswho visited the analysis sample of 68 EDs. Of thiscohort, 166,334 (74.5%) were discharged home from theED, and 42,343 were discharged home following hospi-talization; the resultant sample of 208,677 was used forthe analysis of outpatient ED visits. Of the remainder,2,129 died in the ED, 4,359 died during hospitalization,3,640 were transferred to LTC, and 4,315 were trans-ferred to a different acute care hospital.

MeasurementsThe classification of three ED types is described indetail elsewhere.9 Briefly, type 1 EDs are the most spe-cialized in their staffing and services, but have weaklinkages with community physicians; these are the larg-est EDs, more often located in metropolitan areas. Type2 EDs are moderately specialized, but have weak link-ages with and poor access to home care and commu-nity services. Type 3 EDs are the least specialized, butmost community-oriented, with better links to commu-nity physicians and services; these EDs are locatedmainly in rural and other nonmetropolitan areas

Patient characteristics were measured with data fromthe 16 months prior to the index visit and include

ACADEMIC EMERGENCY MEDICINE • March 2012, Vol. 19, No. 3 • www.aemj.org 305

measures of characteristics associated with greatercomplexity of patient care and worse outcomes amongseniors.3,16–18 Social and demographic characteristics ofpatients included age, sex, and two measures of area ofresidence. Metropolitan influence zone (MIZ) is aseven-level ordinal variable measuring the influence oflarge urban core areas on residence and place of workfrom the highest (areas inside or close to a major urbancore) to the lowest (areas with no major urban coreinfluence).19 We classified MIZ level into three catego-ries: metropolitan (1), other urban (2–4), and rural (5–7).We measured proximity to the ED of the index visit bywhether or not the patient lived in the same local healthservice area where the ED was located. Measures ofhealth status included the diagnosis at the index EDvisit and the Charlson Comorbidity Index (CCI) forICD-9 administrative data, based on hospital dischargeand ambulatory diagnoses.19,20 Health services utiliza-tion measures included numbers of hospital days, out-patient ED visits, and physician visits (family doctor andspecialist visits) during the 16-month baseline period.

We measured two types of outcome during the6 months after the index visit: more serious and less

serious. More serious outcomes included death, hospi-talization (including admission either at the index EDvisit or at a return visit), and LTC admissions (includingpatients discharged to LTC as well as those hospitalizedfor at least 30 days, as most of the latter are awaitingLTC placement).17 We also created a composite mea-sure of serious outcomes—death, hospitalization, orLTC admission. One less serious outcome—the propor-tion with a return outpatient ED visit (i.e., without hos-pital admission)—was measured among patients whowere discharged either from the ED or hospital.

Data AnalysisWe obtained a simple description of the distribution ofpatient characteristics across ED types by calculatingthese distributions for each ED unit and then averagingwithin each ED type. To assess statistical significance ofthese differences, we used a multilevel approach,21 con-sidering that a patient is nested within an ED unit, andED units are nested within ED types. Significance levelswere obtained for each patient characteristic by fittinga generalized linear mixed model (GLMM)22,23 andcomputing the Wald statistic of the ED type variable.

Figure 1. Derivation of patient cohort for analysis. RAMQ = Regie de l’assurance Maladie du Quebec (physician billing database).

306 McCusker et al. • OUTCOMES OF SENIORS AND ED TYPE

A specific GLMM family was used depending on thetype of response: normal (age, number of family doctorand specialist visits); binomial (sex, resident outsideservice area, diagnoses at index ED visit); multinomial(MIZ); and negative binomial (CCI, hospitalization,number of ED visits). Although measures of effect aregenerally preferable as they convey more informationthan p-values, the presentation of effect measureswould have been cumbersome in this situation. There-fore, we present only nominal p-values. We did notadjust for multiple comparisons.

We used Cox regression analysis24,25 to predict theeffect of ED type on the following patient-level (cen-sored) times: 1) time to death, 2) to acute care admis-sion, 3) to LTC admission, 4) to the first serious event(death, acute care, or LTC admission), and 5) to firstoutpatient ED visit. Time zero for these analyses wasthe index ED visit for the first four outcomes(n = 223,120) and discharge from the ED or hospital forthe last outcome (n = 208,677). Censoring was definedas due to either end of study or occurrence of deathwhen death is not part of the outcome under study in aspecific survival model. The Cox models were used tocompute the hazard ratios (HRs) and 95% confidenceintervals (CI). In each case, we checked the proportionalhazard assumption using Schoenfeld test. Potentialpresence of multicollinearity was assessed by calculat-ing the variance inflation factor (VIF) for each variable.

Unadjusted models were fitted for each outcome.Multivariate models were developed to adjust for base-line patient characteristics using the direct variableselection method. To adjust optimally for comorbidityburden, we explored the use of a multivariate comorbid-ity confounder score26 as an alternative to the CCI. Thecomorbidity confounder score was based on the 17 CCIdiagnoses, injuries and mental disorders, and the num-ber of different medications. Because the confounderscore performed similarly to the better-known CCI, wepresent only results adjusted for CCI. Other covariatesthat were included in the multivariate models were age,sex, proximity to ED, urban–rural residence, hospi-tal days, outpatient ED visits, family physician visits, andspecialist visits. Potential ED-level effects were takeninto account using a frailty term: a random effect foreach ED that acts multiplicatively on the baseline haz-ard.27 Survival curves were constructed for each out-come and difference across ED types was tested withthe log-rank test. Finally, to assess whether the effect ofED type differed for certain patient subgroups, wetested for interactions of ED type with patient age, sex,CCI, and urban–rural residence. Interaction terms withp-values less than 0.10 were considered to be statisti-cally significant and the interactions were further evalu-ated in stratified analyses. All calculations were carriedout in SAS 9.2 (PROC NLMIXED, PROC MIXED; SASInstitute, Cary, NC) and STATA 10.0 (STCOX, XTLOGIT,XTNBREG; StataCorp, College Station, TX).

RESULTS

Patient CharacteristicsIn comparison with community-dwelling seniors trea-ted at the least specialized EDs (type 3), those at the

most specialized EDs (type 1) were most likely to live ina metropolitan area, outside the ED health service area(Table 1). At the index visit, they were more likely to bediagnosed with a circulatory disorder, cancer, or gen-eral symptoms and less likely to receive a respiratorydiagnosis. They had higher levels of comorbidity andhigher numbers of prior specialist visits. Patients treatedat type 2 EDs were intermediate between those at type 1and type 3 EDs with respect to these characteristics.

We also compared patient characteristics of the anal-ysis sample with those who visited the 35 EDs excludedfrom the analysis because of incomplete organizationalinformation (data not shown). None of these character-istics differed significantly except for a mental or ner-vous system diagnosis at the index visit (6.6% in thesample vs. 7.3% at nonparticipant ED, p = 0.038).

Patient OutcomesSurvival curves for the study outcomes are shown inFigures 2A to 2D. The log-rank test showed that all dif-ferences across ED types are significant at the 0.05 level(p < 0.001). Risk of death was somewhat greater for allED types during the earlier part of the follow-up (thesecurves include deaths during the index visit and initialhospitalization). Risk of hospitalization was muchgreater for all ED types at the index visit than later inthe follow-up period. The type 1 EDs had higher risksof all these outcomes, but particularly for LTC admis-sion, compared to the type 3 EDs. The type 2 EDs hadrisks of serious outcomes that were intermediatebetween those for types 1 and 3 (the 95% CIs excludedunity for all outcomes except LTC).

The Cox model was used to adjust for patient charac-teristics (Table 2). It should be noted that ED type com-parisons (type 1 vs. type 3 and type 2 vs. type 3) for thefour serious outcomes (mortality, LTC, hospital admis-sion, and the composite) do not satisfy the proportionalhazards assumption (data not shown). For these fouroutcomes, the HR decreased by not more than 1% permonth; CI sizes remained of similar width across time.The contribution of the frailty term was significant(p < 0.001) for all four outcomes; the proportions of EDvariance explained (indicating heterogeneity) were 8,10, 11, and 24% for mortality, serious outcome, hospitaladmission, and LTC, respectively.

In contrast to the results for the serious outcomes,there were fewer outpatient ED return visits (Figure 2E)among patients who visited the most specialized (type1) versus the least specialized (type 3) EDs, with type 2EDs intermediate between the other two ED types (log-rank test, p < 0.001). The 95% CIs of the HRs excludedunity only for the comparison between type 1 and type3 EDs. Note that, among those patients who were dis-charged home at the ED visit, the overall rate of returnED visits ranged from 43.0% in type 1 to 47.3% in type3 EDs (data not shown). The proportion with a returnED visit at which they were hospitalized ranged from31.1% at type 1 to 23.2% at type 3 EDs.

For all the multivariate Cox models shown in Table 2,multicollinearity did not appear to cause serious prob-lems, as the VIFs did not exceed 2. The HRs for each offive outcomes were similar in univariate and multivariateanalyses when the frailty term was used (less than 2%


of change in the effect of ED type, data not shown).However, the addition of the frailty term accounted formost of the change in the HR and had a substantialeffect on the 95% CI, resulting in wider CIs (seeTable 2). There was more heterogeneity in the out-comes within EDs of types 2 and 3 compared to type 1.These models do not include the interaction terms thatare described below.

Significant interactions were found for ED type withage and comorbidity score both for the composite seri-ous outcome measure and for outpatient ED visits; thedifferences in these outcomes by ED type were attenu-ated among older patients and those with higher

comorbidity scores (data not shown). Among patientsaged 85 years and over, the differences between types1 and 3 EDs in both serious outcomes and outpatientED visits were no longer statistically significant (95%CIs included unity). Among patients with higher comor-bidity scores (3+), the differences between both types 1and 2 compared to type 3 EDs for serious outcomeswere attenuated (type 1 HR 1.23 [95% CI = 1.01 to1.48]), whereas among patients with lower comorbidity,the HR for type 1 EDs was 1.39 (95% CI = 1.10 to 1.76).The interaction with outpatient ED visits had only aminor effect on the HR, which remained statisticallysignificant both among those with higher and lower

Table 1Patient Profiles by ED Type

Patient Characteristics

Type 1: MostSpecialized, Less

Community-orientedEDs (%)*

Type 2: ModeratelySpecialized, Less


Type 3: LeastSpecialized, More


Wald Statistics�p-value

DemographicsAge (yr) 0.032

65–74 47.8 50.2 52.175–84 39.3 37.9 37.385+ 13.0 11.9 10.6

Female 56.5 57.2 56.3 0.562Area of residence <0.001

Metropolitan 66.1 52.3 24.4Other urban 18.4 26.5 28.5Rural 15.5 21.1 47.1Resident outside service area 57.3 42.5 22.0 <0.001

Health statusDiagnosis at index visit

Circulatory 19.3 17.6 13.5 <0.001Respiratory 8.7 10.3 12.4 0.001Cancer 2.7 2.4 1.7 <0.001Mental or nervous 6.5 6.8 6.4 0.652Injury 16.6 15.1 16.7 0.239Other diagnosis 30.4 31 30.2 0.506General symptoms 34.3 29.6 23.7 <0.001

CCI <0.0010 40.8 45.1 49.31–2 34.3 32.5 31.63+ 24.8 22.4 19.1

Health services utilization (previous 16 months)Hospitalization 0.227

No 79.3 78.2 79.7<7 days 8.4 8.5 8.37 days or more 12.3 13.3 12.0

Number of outpatient ED visits 0.0130 69.1 62.3 59.21 17.3 19.0 19.92+ 13.6 18.7 20.9

Number of family doctor visits 0.0510 15.7 17.4 22.41–4 32.8 35.1 36.35–9 31 30.8 27.5

10+ 20.6 16.7 14.3Number of specialist visits <0.001

0 16.7 18.7 22.81–4 33.2 37.6 41.65–9 23.7 23.8 21.6

10+ 26.4 19.9 14

Type 1 = 12 EDs and 66,709 patients; type 2 = 28 EDs and 97,852 patients; type 3 = 28 EDs and 58,559 patients.CCI = Charlson Comorbidity Index.*Percentages are computed for each ED (n = 68), and the mean percentage is computed for each ED type.�Based on multilevel models, see Materials and Methods.


comorbidity scores. The interactions of ED type withpatient sex and urban–rural residence were not statisti-cally significant for either outcome.

DISCUSSION

We conducted a comprehensive population-based studyof characteristics and outcomes of community-dwellingseniors who visited three different types of ED, basedon a recently developed classification.9 The main find-ing is that seniors who visited type 1, compared to type3 EDs, were both more vulnerable and experiencedmore serious outcomes, even after adjustment for soci-odemographic factors, comorbidity score, and prior uti-lization of health services. Measures of their greatervulnerability included higher average comorbidity

scores and a high proportion presenting with nonspe-cific complaints, an indicator of increased frailty and arisk factor for poor hospital outcomes.28

There are several plausible explanations of theseresults. First, self-selection may determine to someextent the type of ED visited. Sicker patients, andthose who use more hospital-based and specialist ser-vices, are more likely to use EDs with which they andtheir physicians are connected. Type 1 EDs are larger,often located in metropolitan areas, with a higher con-centration of specialists appointed to these institutions.Second, the worse outcomes at type 1 EDs are likely tobe explained in large part by unmeasured patient char-acteristics (e.g., more severe illness, greater functionaland cognitive impairment); they should be interpretedprimarily as indicators of greater vulnerability. An

Death Long-term care admission

Hospital admission

Return outpatient ED visit

Death, long-term care or hospital admission

Figure 2. Six-month survival curves for each outcome by three ED types. y-axis = proportion without the outcome; x-axis = daysfrom the index visit (A–D) or from discharge (E). (—) Most specialized, less community-oriented EDs (A–D, n = 66,709; E,n = 61,929). (- - -) Moderately specialized, less community-oriented EDs (A–D, n = 97,852; E, n = 91,219). (ÆÆÆ) Least specialized, morecommunity-oriented EDs (A–D, n = 58,559; E, n = 55,529).


alternative interpretation, however, is that these out-comes are due in part to the lower accessibility of homecare services and limited communication with commu-nity physicians.9 Linkages to community services areessential to adequate discharge planning3,5 and for con-tinuity of care between the ED and community.29 Thereare admittedly greater challenges to developing commu-nity linkages in metropolitan areas where type 1 EDs arelocated, as patients come from many different healthservice areas and ED staff have to deal with many moreagencies and services than those in more rural areas.Furthermore, the proportion of the population with afamily doctor is lowest in metropolitan areas of theprovince.10 The closing of family medicine solo practicesand the move to group practices, more frequent inmetropolitan areas of the province, is leaving manyolder patients in these regions without a family doctor.30

Seniors with less complex needs were more likely toreceive care at a more community-oriented (type 3) ED.Furthermore, seniors at type 3 EDs had higher rates ofreturn ED visits overall, and a lower proportion werehospitalized at these return visits, indicating that theseEDs play a greater role in providing primary care.31

Indeed, many of these EDs are located in more rural

areas, where general practitioners are more likely topractice in multiple locations, including the ED.32

This study is, to our knowledge, the first to examineoutcomes of an ED visit among community-dwellingseniors using an ED classification that is based on bothinternal and external aspects of resources and organiza-tion of geriatric services. One prior study that examineda restricted set of ED services found that smaller EDs,with more limited resources, lacking inpatient geriatricservices, or without a social worker, have higher ratesof return ED visits.7 A second study found differencesby ED size in the services available for seniors and in1-month outcomes in a cohort of seniors dischargedhome from these EDs.8 Rather than examining the rela-tionships of individual ED services to patient outcomes,our current study takes a more comprehensiveapproach, assessing ED types more globally, based onmultiple organizational features. The two extreme types(1 and 3) were the most clearly differentiated in thisstudy. Further work may need to examine possible sub-types in the intermediate, type 2, category of EDs,which was more heterogeneous than the two extremeED types.

Some patient characteristics—age and comorbidity—appear to modify the relationship of ED type to out-comes. The association of ED type with more seriousoutcomes was attenuated in the oldest age group (thoseaged 85 years and over) and among those with higherlevels of comorbidity. One plausible interpretation ofthese results is a selection effect, whereby seniors withmore severe illness are more likely to be referred to amore specialized ED if they are younger than 75 yearsand have less complex needs. Potentially life-saving andlife-enhancing investigations and interventions tend tobe used less frequently in older populations.33,34 As nomeasure of acuity was available in this study, furtherinvestigation was not possible.

This study underlines the need to distinguish outpa-tient ED visits (those at which patients are not hospital-ized) from more serious outcomes. Althoughresearchers sometimes combine these outcomes into acomposite outcome variable, there are several reasonsto differentiate these two types of outcome. First, EDvisits for less serious problems are more likely to sub-stitute for community-based primary care.11 Second,frailty predicts serious outcomes but not outpatient EDvisits.35 Third, ED-based geriatric interventions that areeffective in reducing serious outcomes appear to begenerally ineffective in reducing outpatient ED visits.36

There are at several implications of the study forfuture research. First, the results of geriatric researchconducted at large, urban EDs may not be generalizableto other types of EDs, given the differences in staffing,services, and patient populations. Second, the extent towhich different types of ED meet the needs of seniorsshould be examined systematically, using a broaderrange of outcomes than those used in this study (e.g.,patient and family perceptions and experiences of thecare they received can help to improve ED services). Inany population, there may need to be a mix of differenttypes of ED to meet the needs of different subgroups ofseniors. Third, we show that more serious outcomesshould be examined separately from outpatient ED

Table 2Six -month Outcomes by ED Type (68 EDs, 223,120 Patients) forCox Models Without Interactions*

Outcome� andED Type Univariate� Multivariate§

MortalityType 1 1.28|| (1.23–1.33) 1.36|| (1.11–1.66)Type 2 1.15|| (1.11–1.19) 1.18– (1.01–1.38)Type 3 1.00 1.00

LTCType 1 1.52|| (1.44–1.61) 1.70|| (1.21–2.41)Type 2 1.26|| (1.19–1.33) 1.19 (0.91–1.56)Type 3 1.00 1.00

Hospital admissionType 1 1.18|| (1.16–1.20) 1.32– (1.04–1.66)Type 2 1.21|| (1.19–1.23) 1.23– (1.03–1.48)Type 3 1.00 1.00

Serious outcome**Type 1 1.20|| (1.17–1.22) 1.32– (1.07–1.63)Type 2 1.21|| (1.19–1.23) 1.22– (1.04–1.44)Type 3 1.00 1.00

ED outpatient visitType 1 0.87|| (0.85–0.88) 0.82|| (0.71–0.93)Type 2 0.90|| (0.88–0.91) 0.91 (0.82–1.01)Type 3 1.00 1.00

Data are reported as HR. (95% CI).HR = hazard ratio; LTC = long-term care.*Please see text for description of models with interactions.�Outcomes are measured from date of index visit for mortal-ity, LTC, hospital admission and serious outcomes(n = 223,120) and from date of ED or hospital discharge foroutpatient ED visits (n = 208,677).�Models with no adjustment and no frailty term.§Models adjusted for age, sex, area of residence, proximityto ED, comorbidity, health services utilization in previous16 months (hospital, ED, family doctor, and specialist visits)and frailty term.||p < 0.01.–0.01 £ p < 0.05.**Death or admission to acute or LTC.


visits, not combined in composite measures of adverseoutcomes. Finally, this study highlights important dif-ferences in serious outcomes between older users ofdifferent types of ED, even after adjustment for patientcharacteristics. It remains to be determined whetherthese differences are due to ED-related factors (e.g., dif-ferences resources and care processes) or to unmea-sured differences in patient populations.

LIMITATIONS

Limitations of the study include those inherent inadministrative databases: questionable reliability of dataand limited information on clinical status. Although weused multiple measures of comorbidity, residual con-founding by unmeasured factors such as disease sever-ity and acuity, and functional and cognitive status,likely accounted in part for the worse outcomes in type1 and 2 EDs. Second, the ED classification we used wasbased primarily on unvalidated data provided by keyinformants on ED services and their perceptions of theavailability of community services. Although these keyinformants were well placed to provide such data,future research might consider using additional infor-mants, both in the ED and in the community. Third, webased our study on those two-thirds of EDs with com-plete organizational information, potentially biasing theresults. However, the excluded EDs did not differ fromthose comprising the study sample either in measuredorganizational or in patient characteristics. Fourth, con-textual factors may affect some of the results of thisstudy (e.g., accessibility of alternative sources ofcare, availability of LTC beds in the community).Finally, the ED types identified in this Quebec studymay not represent those in other locations or othercountries.

CONCLUSIONS

More vulnerable community-dwelling seniors tend tobe treated in more specialized EDs, which have poorerlinkages to community services. Improved linkagesbetween more specialized EDs and the community(physicians, home care, and other services) andincreased access to community services may improveoutcomes in this population. Increased rates of outpa-tient ED return visits at less specialized, more commu-nity-oriented EDs reflect the greater role of these EDsin providing primary medical care.

References

1. Aminzadeh F, Dalziel WB. Older adults in the emer-gency department: a systematic review of patternsof use, adverse outcomes, and effectiveness of inter-ventions. Ann Emerg Med. 2002; 39:238–47.

2. McCusker J, Verdon J, Tousignant P, Poulin deCourval L, Dendukuri N, Belzile E. Rapid emergencydepartment intervention for elders reduces risk offunctional decline: results of a multi-center random-ized trial. J Am Geriatr Soc. 2001; 49:1272–81.

3. Hastings SN, Heflin MT. A systematic review ofinterventions to improve outcomes for elders

discharged from the emergency department. AcadEmerg Med. 2005; 12:978–86.

4. McCusker J, Jacobs P, Dendukuri N, Latimer E,Tousignant P, Verdon J. Cost-effectiveness of abrief 2-stage emergency department interventionfor high risk elders: results of a quasi-randomizedcontrolled trial. Ann Emerg Med. 2003; 41:45–56.

5. Sinha SK, Bessman ES, Flomenbaum N, Leff B. Asystematic review and qualitative analysis to informthe development of a new emergency department-based geriatric case management model. AnnEmerg Med. 2011; 57:672–82.

6. Samaras N, Chevalley T, Samaras D, Gold G. Olderpatients in the emergency department: a review.Ann Emerg Med. 2010; 56:261–9.

7. McCusker J, Ionescu-Ittu R, Ciampi A, et al. Hospi-tal characteristics and emergency department careof older patients are associated with return visits.Acad Emerg Med. 2007; 14:426–33.

8. McCusker J, Roberge D, Vadeboncoeur A, VerdonJ. Safety of discharge of seniors from the emer-gency department to the community. Healthc Q.2009; 12:24–32.

9. Borgès Da Silva R, McCusker J, Roberge D, et al.Classification of emergency departments withrespect to their services for community-dwellingseniors. Acad Emerg Med. 2012; 19: in press.

10. Sibley L, Weiner J. An evaluation of access to healthcare services along the rural-urban continuum inCanada. BMC Health Serv Res. 2011; 11(1):e20.

11. Petersen LA, Burstin HR, O’Neil AC, Orav EJ, Bren-nan TA. Nonurgent emergency department visits:the effect of having a regular doctor. Med Care.1998; 36:1249–55.

12. Schoen C, Osborn R, Huynh P, et al. Taking thepulse of health care systems: experiences of patientswith health problems in six countries. Health Aff.2005:509–25.

13. Schoen C, Osborn R, Doty MM, Bishop M, Peugh J,Murukutla N. Toward higher-performance healthsystems: adults’ health care experiences in sevencountries. Health Aff (Millwood). 2007; 26:717–34.

14. Roberge D, Pineault R, Larouche D, Poirier LR. Thecontinuing saga of emergency room overcrowding:are we aiming at the right target? Healthc Policy.2010; 5:27–39.

15. Belzile E, Sanche S, McCusker J, Vadeboncoeur A,Ciampi A, Levesque JF. A Measure of EmergencyDepartment Use Based on Quebec’s AdministrativeData. Available at: http://smhc.qc.ca/ignitionweb/data/media_centre_files/490/A%20Measure%20of%20Emergency...%20EN_May%202011.pdf. AccessedDec 7, 2011.

16. Hastings NS, Schmader KE, Sloane RJ, WeinbergerM, Goldberg KC, Oddone EZ. Adverse health out-comes after discharge from the emergency depart-ment–incidence and risk factors in a veteranpopulation. J Gen Intern Med. 2007; 22:1527–31.

17. McCusker J, Bellavance F, Cardin S, Trepanier S,Verdon J, Ardman O. Detection of older people atincreased risk of adverse health outcomes after anemergency visit: the ISAR screening tool. J AmGeriatr Soc. 1999; 47:1229–37.


18. Hastings NS, Oddone EZ, Fillenbaum G, Sloane RJ,Schmader KE. Frequency and predictors of adversehealth outcomes in older medicare beneficiaries dis-charged from the emergency department. MedCare. 2008; 46:771–7.

19. McNiven C, Puderer H, Janes D. Census Metropoli-tan Area and Census Agglomeration InfluenceZones (MIZ): A Description of the Methodology.Ottawa. 2000. Available at: http://www.statcan.gc.ca/bsolc/olc-cel/olc-cel?lang=eng&catno=92F0138M2000002. Accessed Dec 11 2011.

20. D’Hoore W, Bouckaert A, Tilquin C. Practical con-siderations on the use of the Charlson ComorbidityIndex with administrative data bases. J Clin Epi-demiol. 1996; 49:1429–33.

21. Snijders T, Bosker R. Multilevel Analysis. An Intro-duction to Basic and Advanced Multilevel Model-ing. London, England: Sage Publications, 1999.

22. Verbeke G, Molenberghs G. Linear Mixed Modelsfor Longitudinal Data. New York, NY: Springer, 2000.

23. Verbeke G, Molenberghs G. Models for DiscreteLongitudinal Data. New York, NY: Springer, 2005.

24. Harrell FE Jr. Regression Modeling Strategies withApplications to Linear Models, Logistic Regression,and Survival Analysis. New York, NY: Springer,2001.

25. Allison PD. Survival Analysis Using SAS. A Practi-cal Guide. Cary, NC: SAS Institute Inc., 1995.

26. Miettinen OS. Stratification by a multivariate con-founder score. Am J Epidemiol. 1976; 104:609–20.

27. Cleves M, Gould W, Gutierrez R, Marchenko Y. AnIntroduction to Survival Analysis Using STATA.College Station, TX: Stata Press, 2008.

28. Jarrett PG, Rockwood K, Carver D, Stolee P,Cosway S. Illness presentation in elderly patients.Arch Intern Med. 1995; 155:1060–4.

29. McCusker J, Dendukuri N, Tousignant P, Verdon J,Poulin de Courval L, Belzile E. Rapid two-stageemergency department intervention for seniors:impact on continuity of care. Acad Emerg Med.2003; 10:233–43.

30. Pineault R, Levesque J, Roberge D, Hamel M, Lam-arche P, Haggerty J. Accessibility and Continuity ofCare: A Study of Primary Healthcare in Québec.Montreal. Available at: http://www.inspq.qc.ca/pdf/publications/911_ServicesPremLigneANGLAIS.pdf.Accessed Dec 7, 2011.

31. Haggerty JL, Roberge D, Pineault R, Larouche D,Touati N. Features of primary healthcare clinicsassociated with patients’ utilization of emergencyrooms: urban-rural differences. Healthc Policy.2007; 3:72–85.

32. Borgès Da Silva R. The Medical Practice of GeneralPractitioners: Influence of Organizational and Geo-graphical Contexts. Montreal, Quebec: University ofMontréal, 2010.

33. Bowling A. Ageism in cardiology. Br Med J. 1999;319:1353–5.

34. Magid DJ, Masoudi FA, Vinson DR, et al. Olderemergency department patients with acute myocar-dial infarction receive lower quality of care thanyounger patients. Ann Emerg Med. 2005; 46:14–21.

35. Hastings SN, Purser JL, Johnson KS, Sloane RJ,Whitson HE. Frailty predicts some but not alladverse outcomes in older adults discharged fromthe emergency department. J Am Geriatr Soc. 2008;56:1651–7.

36. McCusker J, Verdon J. Do geriatric interventionsreduce emergency department visits? A systematicreview. J Gerontol A Biol Sci Med Sci. 2006; 61:53–62.


Documents

Outcomes of Community-dwelling Seniors Vary by Type of Emergency Department