A Population-based Study of the Association Between Socioeconomic Status and Emergency Department Utilization in Ontario, Canada

A Population-based Study of the AssociationBetween Socioeconomic Status andEmergency Department Utilizationin Ontario, CanadaYasmin Khan, MD, MPH, Richard H. Glazier, MD, MPH, Rahim Moineddin, PhD,and Michael J. Schull, MD, MSc

AbstractObjectives: The relative effects of socioeconomic status (SES) and health status on emergencydepartment (ED) utilization are controversial. The authors examined this in a setting with universalhealth coverage.

Methods: For Ontario participants age 20–74 years, Canadian Community Health Survey 2000 to 2001responses were linked to Ontario Health Insurance Plan (OHIP) physician utilization data for 1999 to2001 and the National Ambulatory Care Reporting System (NACRS) for ED utilization in 2002. SES wasdefined primarily according to high school completion and secondarily according to income. The pri-mary outcome was less urgent ED visit, defined as Canadian Triage and Acuity Scale (CTAS) 4 or 5 andnot admitted to hospital.

Results: The weighted sample was 9,323,217. Overall, 31.4% of the sample used an Ontario ED in 2002.The majority of visits (59.1%) were classified as less urgent. Fair or poor self-perceived health was thelargest predictor of ED use, regardless of visit urgency. Respondents with low education were morelikely to have both less urgent visits (odds ratio [OR] = 1.65, 95% confidence interval [CI] = 1.35 to 1.94)and more urgent visits (OR = 1.39, 95% CI = 1.09 to 1.68) after controlling for age, sex, income, self-perceived health, urban or rural location, regular doctor, and non-ED physician visits. Education wasnot associated with having less urgent versus more urgent visits (OR = 0.92, 95% CI = 0.68 to 1.14).

Conclusions: In a setting with universal health insurance, worse health status is the largest predictor ofED utilization, but low SES is independently associated with increased use of the ED, regardless of visiturgency. This study lends support to findings in other health systems that those using EDs are more illand more disadvantaged.

ACADEMIC EMERGENCY MEDICINE 2011; 18:836–843 ª 2011 by the Society for Academic EmergencyMedicine

E mergency department (ED) use continues to risein North America, and worsening crowdingcompromises patient care.1,2 Understanding

contributing factors and reasons for use, therefore, hashealth policy relevance. In the United States, previousstudies conducted at single sites indicated that thosewho use the ED were more likely to be the uninsured,patients without primary care physicians, visible minor-ities, and other ‘‘vulnerable populations.’’3–6 However,recent studies that examined population-based samplesin the United States found that those using EDs weresimply more likely to be in poor health and to haveexperienced disruptions in care.7,8 Canada’s health sys-tem differs from that of the United States in that healthinsurance is universal; thus, access to health careshould not vary by age, income, or other factor, afteraccounting for health needs.9 Assuming equal access tohealth care, use of EDs should theoretically not vary by

ISSN 1069-6563 ª 2011 by the Society for Academic Emergency Medicine836 PII ISSN 1069-6563583 doi: 10.1111/j.1553-2712.2011.01127.x

From the Division of Emergency Medicine, Department of Medi-cine (YK, MJS), the Dalla Lana School of Public Health (YK, RHG,RM), and the Department of Family & Community Medicine(RHG, RM), University of Toronto, Toronto, Ontario; the Institutefor Clinical Evaluative Sciences (RHG, RM, MJS), Toronto,Ontario; the Centre for Research on Inner City Health,St. Michael’s Hospital (RHG), Toronto, Ontario; and the Depart-ment of Emergency Services and Clinical Epidemiology Unit,Sunnybrook Health Sciences Centre (MJS), Toronto, Ontario,Canada.Received December 8, 2010; revisions received February 3and February 7, 2011; accepted February 7, 2011.Presented at the Canadian Association of Emergency Physiciansannual conference, Victoria, British Columbia, Canada, June 2007.The authors have no disclosures or conflicts of interest to report.Supervising Editor: Lowell Gerson, PhD.Address for correspondence and reprints: Yasmin Khan, MD,MPH; e-mail: [email protected].

socioeconomic status (SES), especially for low urgencyconditions that potentially can be managed in non-EDsettings. There is no population-based informationfrom Canada to document the characteristics of thosewho use the ED compared with those who do not andwhether the principles of Canadian universal healthinsurance hold true.

Examining waiting times for health services is cur-rently a priority for the Canadian health system, incitedby the 2003 First Ministers’ Accord.10 A 2005 report bythe Canadian Institute for Health Information investi-gating ED wait times found that 57% of ED visits arefor less urgent or nonurgent reasons.1 An Ontario-based report found a lack of variation in use of primarycare and specialist physician services for disadvantagedgroups compared with nondisadvantaged groups,despite worse health status.11 The definition of disad-vantaged used in this report included SES, unmethealth needs, rural residence, immigration, and ethnic-ity. The report also demonstrated higher frequencies ofED use specifically for those of lower SES. These trendswere hypothesized to represent a barrier that disadvan-taged populations, particularly those with low SES,experience in accessing a universal health care system,which results in increased use of EDs.

Lower SES is understood to result in worse healthstatus, which predictably leads to increased health careutilization.12 In addition to this, individuals of lowerSES may be more likely to encounter barriers in navi-gating the health care system.11,13 This could result inproportionately more ED visits than those with higherSES, with the ED representing the ‘‘safety net’’ whenother providers are inaccessible. Specifically, withincreasing shortages of family physicians, those oflower SES may be less likely to obtain the care of afamily physician. They may, therefore, have to turn tothe ED for care, especially for less urgent or lower-acuity reasons, although these can often be managed innon-ED settings.14

The purpose of this study was to examine a popula-tion-based sample from Ontario, Canada, for variationof ED utilization across socioeconomic groups. Theprimary outcome measure was less urgent ED use, andthe secondary outcomes were overall ED utilizationand more urgent utilization. Our hypothesis wasthat individuals with lower SES would be more likelyto have increased ED utilization for less urgentconditions.

METHODS

Study DesignThis was a secondary analysis of data from the Cana-dian Community Health Survey (CCHS)–Cycle 1.1. Thestudy was approved by the institutional research ethicsboard at Sunnybrook Health Sciences Center.

Study Setting and PopulationThe study setting of Ontario is Canada’s largest prov-ince with almost 12 million people. The target popula-tion of the CCHS included household residents in allprovinces and territories, with the principal exclusionof populations in First Nations reserves, Canadian

Forces Bases, and some remote areas. The CCHSemployed a multistage stratified cluster design and theOntario portion consisted of 37,681 respondents.

Our study population was restricted to individualsbetween the ages of 20 and 74 years to avoid proxyresponses that could be assigned to children and olderseniors. The sample was stratified by sex and age fordescriptive purposes.

Study ProtocolThe CCHS was a national survey conducted by Statis-tics Canada in 2000–2001. It was designed to providetimely cross-sectional estimates of health determinants,health status, and health system utilization at a subpro-vincial level (health region or combination of healthregions). CCHS respondents were asked to providetheir Ontario health card numbers and to consent tolinkage with their health care utilization data. Thoseconsenting were linked to the Ontario Registered Per-sons Database (RPDB), the province’s health care regis-try. Once linked with the RPDB, health card numberswere used to link respondents with physician fee-for-service claims to the Ontario Health Insurance Plan(OHIP) for 2000–2001, which corresponded to one fiscalyear (approximately 94% of all physician encounters inthe province are included in this database) and torecords of ED visits for 2002 from the National Ambula-tory Care Reporting System (NACRS), for which closeto 100% of ED claims in the province are included. Theyear 2002 was chosen for NACRS data since 2002 wasthe first largely complete year for the database. Thedata were accessed as part of a comprehensiveresearch agreement with the Ontario Ministry of Healthand Long-Term Care.

Definitions. Socioeconomic status as a measure of dis-advantage can be conceptualized in different ways. Forthe purposes of this study, the main predictor variableswere as follows. Low education was the primary inde-pendent variable and was defined as not having gradu-ated from high school. Low income was viewed as animportant predictor of SES and was defined by theCCHS categories of lowest, lower middle, and middleincome. Income, however, was viewed as a secondaryindependent variable due to the potential for missingvalues. The lowest, lower middle and middle incomecategories were defined as a household income of£$29,999, £$39,999, or £$59,999, respectively.

Covariates from the CCHS for inclusion in the analy-sis were determined a priori. The CCHS variable ofself-perceived health was defined as the primary indica-tor of health need. This variable has been validated andreliably found to represent an individual’s healthneeds.15,16 Living in a rural area as defined by theCCHS was also considered a health need variable, dueto the reduced availability of some health care servicesand providers in rural areas of Ontario.11

Use of non–emergency physician health services wasincluded as an additional measure of health care needas well as overall health care-seeking behavior. Thiswas measured for family physician ⁄ general practitioner(FP ⁄ GP) visits, as well as specialist visits based on OHIPhealth insurance claims. The CCHS variable of ‘‘regular

ACADEMIC EMERGENCY MEDICINE • August 2011, Vol. 18, No. 8 • www.aemj.org 837

medical doctor’’ was used to control for having aregular source of care.

Outcome MeasuresThe primary outcome measure was less urgent visits toan ED. Less urgent was defined according to the Cana-dian Triage and Acuity Scale (CTAS) values of 4 (lessurgent) and 5 (nonurgent) and excluded those whowere admitted to hospital.17 Secondary outcome mea-sures were 1) more urgent ED visits, defined as CTAS1, 2, and 3 (resuscitation, emergent, and urgent, respec-tively) and including all patients admitted to hospitalfrom any triage category; and 2) ED visits of any triagecategory. For those with repeat visits to EDs in thestudy year, we included only the first ED visit in ouranalysis; thus, demographic characteristics were ana-lyzed once per individual. CTAS values were obtainedfrom the NACRS database of ED records and are basedon the score assigned by a triage nurse at the time ofED triage.

Data AnalysisRegression analysis was used to test the associationbetween the independent variables and the outcomes,while controlling for the potential confoundersdescribed above. All covariates of interest were enteredinto the model simultaneously. Multivariate logisticregression was used to fit the model for the dichoto-mous outcomes. The dichotomous outcomes used weremore urgent, less urgent, and all ED visits, each sepa-rately compared with the group that had no ED visits.Last, more urgent and less urgent visit groups werecompared with each other. Poisson regression wasused to model the number of ED visits. Statistical sig-nificance was defined at the 0.05 level. The analysis wasconducted using SAS Version 9.1 (SAS Institute Inc.,Cary, NC).

All analyses were performed with the weight variableprovided by Statistics Canada to reflect samplingdesign and variation.18 Unweighted counts of less than30 are not reported. Estimates with a coefficient of vari-ation of >33.3% are not reported, while those between16.5 and 33.3% are reported as marginal estimates tobe interpreted accordingly, as recommended by Statis-tics Canada.18 All confidence intervals (CIs) and thecoefficients of variation were calculated by means ofbootstrap methods with 500 replications using boot-strap weights provided by Statistics Canada. Model fitfor regression models was assessed using the ratio ofdeviance to degrees of freedom, with good fit beingclose to 1.

Sensitivity Analyses. Visits to EDs were examined bydiagnostic category for the comparison groups accord-ing to the major categories used by the InternationalClassification of Diseases revision 10 (ICD-10).19 Thiswas done by age strata of 20 to 34, 35 to 49, 50 to 64,and 65 to 74 years to address possible confounding byage. In addition, day of the week (weekend or holidayversus weekday) and time of day (daytime defined as08:00 to 20:00 and after hours as 20:00 to 08:00) wereexamined to account for any differences in times forseeking care that could occur between comparison

groups. CTAS 3 is an intermediate category, and forthat reason we repeated primary analyses using solelyCTAS 1 and 2 as more urgent and CTAS 3, 4, and 5 asless urgent.

It was anticipated that some aspects of health needmay not be captured by a combination of self-perceivedhealth and non–emergency physician visits. For thisreason, additional covariates from the CCHS of self-reported disability, depression, having two or morechronic diseases, and unmet health needs in the past12 months, were considered a priori as secondaryindicators of health need.

The Johns Hopkins Adjusted Clinical Group (ACG)Case-Mix System was also used to account for patientcomorbidity.20 This method of adjusting for case-mixhas previously been used in Canada, in addition tohaving been validated in the United States.21–23 Thesystem uses individual-level data to assign measures ofresource use and comorbidity from diagnoses duringa specified time period, obtained from patient records.In this study, OHIP health insurance claims and hospi-tal admissions from 2001 were used to determinecase-mix using the ACG software.20 The variablesfrom the ACG software that were input into the modelwere adjusted clinical groups and resource utilitybands.

Finally, it was hypothesized that education andincome could interact with self-perceived health andthat this may vary by sex or age. Consequently, educa-tion and income were tested for an interaction withself-perceived health and examined for both sexes andby age category.

RESULTS

Characteristics of Study SubjectsIn Ontario, 87.2% of CCHS respondents provided theirhealth card numbers, consented to the linkage, andwere successfully linked to the RPDB. The weightedsample of study subjects of ages 20 to 74 yearsthat were linked by health card represented 9,323,217Ontarians. The largest proportion of missing values(10.3%) occurred for the income variable. The meanage of the overall sample was 46.0 years (SD±14.7 years) and the proportion of males was 46.3%overall.

Characteristics of the study subjects are shown inTable 1. According to Statistics Canada policy, onlynationally weighted population estimates are provided.Overall, 31.4% of the sample used an Ontario ED in2002. An estimated 59.1% of visits were classifiedaccording to CTAS as less urgent or nonurgent. Of allED visits, approximately 40% occurred during day-time hours Monday to Friday (excluding holidays).

Regarding utilization of other sources of health careservices, 90.9% reported having a regular medical doc-tor and 67.8% had more than four visits to a FP ⁄ GP.More than 50% of individuals also had more than fourvisits to a specialist physician.

Individuals with low education and low income weremore likely to visit the ED at least once (rate ratios[RRs] = 1.37 and 1.30, respectively; Table 2) and evenmore likely to have multiple visits (RR = 1.69 and 1.62,

838 Khan et al. • ASSOCIATION BETWEEN SOCIOECONOMIC STATUS AND ED UTILIZATION

respectively). There was no statistically significantdifference in ED triage score by education or income.Those living in a rural area were more likely to visitEDs, but had visits of lower urgency. Individuals withgreater self-reported health needs, including fair orpoor health, disabilities, and multiple chronic diseases,were more likely to visit the ED (RR = 1.74, 1.54, and1.44, respectively) and their visits were of higherurgency. The results were similar whether more-urgentvisits were defined as CTAS 1 and 2 or as CTAS 1, 2,and 3 (results for CTAS 1 and 2 not shown). Goodnessof fit was adequate with the ratio of deviance todegrees of freedom equaling 0.98.

Regression Analysis: Less Urgent Visits ComparedWith No ED VisitsLow education was independently associated with ahigher risk of having a less urgent ED visit comparedwith having no ED visits (adjusted odds ratio[AOR] = 1.65, 95% CI = 1.35 to 1.94; Table 3). Individuals

with fair or poor self-perceived health had the highestindependent association with risk of using the ED forless urgent presentations (AOR = 1.80, 95% CI = 1.41 to2.19). Having a regular medical doctor was associatedwith decreased risk of using the ED (AOR = 0.75, 95%CI = 0.56 to 0.93). These trends persisted in a model thatexamined those with multiple ED visits (data notshown). The ratio of deviance to degrees of freedomwas 1.5.

Regression Analysis: More Urgent Visits ComparedWith No ED VisitsLow education also had a significantly higher risk ofmaking a more urgent ED visit (AOR = 1.39, 95%CI = 1.09 to 1.68). Age, geographic location, and havinga regular doctor were not significantly associated withincreased risk for more urgent visits. Fair or poor self-perceived health remained the single largest predictorof risk for making an ED visit. The ratio of deviance todegrees of freedom was 1.5.

Table 1Characteristics of CCHS Respondents and Their ED and Other Health Care Utilization in 2002

Characteristic N = 9,323,317*

Males (%) Females (%)

20–49 yrs(n = 3,088,763)

50–74 yrs(n = 1,500,347)

20–49 yrs(n = 3,132,362)

50–74 yrs(n = 1,601,845)

SES indicator�Low education� 18.7 14.0 29.8 11.5 31.1Low income§ 34.6 31.0 34.0 34.3 42.7

Health indicator�Self-rated health fair or poor 14.3 9.0 24.0 10.3 22.8Disability 14.5 6.6 18.5 11.9 31.0Depression 10.5 7.6 5.4 15.7 10.7Two or more chronic diseases 43.1 28.9 52.6 40.2 67.4Has regular medical doctor 90.9 84.2 95.5 92.2 97.1Rural residence 15.2 14.7 17.5 14.6 15.0

ED utilization||ED visits in 2002 31.4 30.8 32.0 31.3 32.2CTAS– 1–2 8.7 5.8 14.8 6.1** 13.3**CTAS 3 32.2 25.3 37.9 33.4 37.6CTAS 4 41.5 45.4 31.9 45.5 36.0CTAS 5 17.6 23.6 15.5 15.1 13.1Weekend visits 30.3 31.7 29.1 29.4 30.6After-hours visits�� 30.8 33.0 25.9 32.1 29.1

Other health care utilization��FP ⁄ GP

No visits 3.7 5.7 3.9 2.2 3.1One to four visits 28.5 44.5 21.4 24.4 14.3More than four visits 67.8 49.8 74.7 73.4 82.6

SpecialistNo visits 10.3 18.3 5.9 9.2 2.3One to four visits 36.9 46.9 29.2 38.2 23.9More than four visits 52.7 34.8 64.9 52.6 73.9

All values are percentages.CCHS = Canadian Community Health Survey; CTAS = Canadian Triage and Acuity Scale; FP ⁄ GP = family physician ⁄ generalpractitioner; NACRS = National Ambulatory Care Reporting System; OHIP = Ontario Health Insurance Plan; SES = socioeconomicstatus.*Representing nationally weighted sample.�Obtained from CCHS self-reported data.�Low education defined as not having completed secondary school.§Low income defined according to CCHS categories of lowest, lower middle, and middle income.||Obtained from NACRS for ED visits.–CTAS definition: 1 = emergent, 2 = urgent, 3 = semiurgent, 4 = less urgent; and 5 = nonurgent.**Marginal estimate with coefficient of variation between 16.5 and 33.3%.��Includes after hours (20:00—08:00 hr) that occur on weekends.��Obtained from OHIP claims.


Regression Analysis: More Urgent Compared WithLess Urgent ED VisitsLow education was not associated with an increasedrisk of having a more urgent visit compared with a less

urgent visit (AOR = 0.92, 95% CI = 0.68 to 1.14). Fair orpoor self-perceived health was, however, associatedwith increased risk of a more-urgent ED visit(AOR = 1.51, 95% CI = 1.07 to 1.95). Rural residence

Table 2Frequency of ED Visits by SES, Location and Health Needs Characteristics

Any ED Visit vs.No Visits

More Than One ED Visitvs. No Visits

More Urgent* Triagefor One or More

ED Visits

% RR� (95% CI) % RR� (95% CI) % RR� (95% CI)

EducationLow 26.8 1.37 11.1 1.69 41.9 1.09High 19.5 (1.25–1.49) 6.6 (1.45–1.93) 38.6 (0.98–1.20)

IncomeLow 24.7 1.30 9.9 1.62 40.1 1.03High 18.9 (1.21–1.39) 6.1 (1.41–1.84) 38.9 (0.93–1.13)

Rural residenceYes 24.7 1.23 8.6 1.21 31.6 0.77No 20.1 (1.13–1.32) 7.1 (1.05–1.37) 41.0 (0.68–0.85)

Self-perceived healthFair or poor 33.3 1.74 16.1 2.59 51.5 1.41Good to excellent 19.2 (1.60–1.88) 6.2 (2.25–2.93) 36.6 (1.26–1.56)

DisabilityYes 30.0 1.54 13.3 2.06 47.8 1.27No 19.5 (1.41–1.67) 6.5 (1.77–2.35) 37.5 (1.13–1.41)

DepressionYes 26.0 1.30 10.8 1.56 39.7 1.02No 20.0 (1.17–1.42) 6.9 (1.30–1.81) 38.9 (0.88–1.17)

‡2 chronic diseasesYes 25.4 1.44 10.3 1.91 44.6 1.30No 17.6 (1.34–1.54) 5.4 (1.67–2.16) 34.3 (1.16–1.44)

Low education = did not complete high school; the lowest, lower middle, and middle income categories were defined as ahousehold income of £$29,999, £$39,999, or £ $59,999, respectively.CTAS = Canadian Triage and Acuity Scale; RR = rate ratio; SES = socioeconomic status.*More urgent in this study defined as CTAS 1–3 based on first visit to an ED in 2002.�Comparing low SES to high SES and greater health needs to lower health needs.

Table 3Logistic Regression Modeling of ED Visit Outcomes

Variable

AORs for ED Visits (95% CI)

Less Urgent* vs.No ED Visit

More Urgent� vs.No ED Visit

ED Visit of Any Tri-age Category vs.

No Visit

More Urgent vs.Less Urgent*

ED Visit

Age (years) 0.97 (0.97–0.98) 1.00 (0.99–1.00) 0.98 (0.98–0.99) 1.02 (1.01–1.03)Males 1.33 (1.14–1.52) 1.15 (0.94–1.36) 1.25 (1.11–1.40) 0.87 (0.67–1.07)Low education 1.65 (1.35–1.94) 1.39 (1.09–1.68) 1.53 (1.30–1.76) 0.92 (0.68–1.14)Low income 1.42 (1.23–1.62) 1.36 (1.10–1.62) 1.38 (1.21–1.56) 0.93 (0.73–1.13)Self-perceived health fair or poor 1.80 (1.41–2.19) 2.78 (2.22–3.33) 2.25 (1.91–2.58) 1.51 (1.07–1.95)Rural vs. urban 1.72 (1.47–2.04) 1.02 (0.83–1.19) 1.42 (1.25–1.64) 0.62 (0.50–0.81)Regular medical doctor 0.75 (0.56–0.93) 0.97 (0.65–1.28) 0.81 (0.65–0.98) 1.39 (0.84–1.94)FP ⁄ GP visits

0 vs. >4 0.39 (0.26–0.52) 0.31 (0.18–0.43) 0.36 (0.26–0.46) 0.84 (0.39–1.28)1–4 vs. >4 0.63 (0.52–0.73) 0.49 (0.40–0.59) 0.58 (0.50–0.66) 0.81 (0.60–1.03)

Visits to specialists0 vs. >4 0.50 (0.38–0.61) 0.31 (0.21–0.40) 0.42 (0.34–0.50) 0.62 (0.37–0.87)1–4 vs. >4 0.67 (0.58–0.77) 0.51 (0.42–0.60) 0.60 (0.53–0.67) 0.79 (0.60–0.97)

AORs = adjusted odds ratios; CTAS = Canadian Triage and Acuity Scale; FP ⁄ GP = family physician ⁄ general physician; low educa-tion = did not complete high school; the lowest, lower middle, and middle income categories were defined as a householdincome of £$29,999, £$39,999, or £$59,999, respectively.*Less urgent in this study defined as CTAS 4 or 5 ED visit not resulting in admission to hospital.�More urgent in this study defined as CTAS 1–3 based on first visit to an ED or admitted patient in 2002.


was associated with less risk of having a more urgentvisit (AOR = 0.62, 95% CI = 0.50 to 0.81). The risk of anED visit was lower among those making fewer FP ⁄ GPand specialist visits, but this effect was not significantfor FP ⁄ GP visits for the comparison of more urgent toless urgent visits. The ratio of deviance to degrees offreedom was 2.2.

Graphs of self-perceived health by education areshown in Figures 1 and 2 for males and females,respectively. The figures demonstrate higher ED useamong those with low education, a finding that is morepronounced among those with better self-reportedhealth and among women. Those with the pooresthealth status have the highest ED use, regardless ofeducation level.

Sensitivity AnalysesWith the persistent association of low education andincome with an increased risk of both low-urgency andhigh-urgency ED, the additional predetermined indica-tors of health need were added to the models: self-reported disability, depression, two or more chronicdiseases, and unmet health needs in the past 12 months.After these variables were added into the model, loweducation continued to be significantly associated witha higher risk of visiting the ED. Similarly, adjusting forindividual-level comorbidity using the ACG Case-Mixsystem did not appreciably alter the strength of theassociation between low education and ED visits. Therewere no statistically significant differences betweeneducation or income groups for weekend or after-hoursvisits.

DISCUSSION

We examined a population-based sample from Ontario,Canada, and found that low education and income,conceptual measures of SES, were independentlyassociated with a modest increase in risk of both high-urgency and low-urgency ED visits. As expected, worsehealth status was the largest determinant of ED useacross all acuity levels, while having a regular medicaldoctor was associated with reduced use of the ED forless urgent visits. Having no contact with non-EDhealth care providers was not associated with increasedED utilization.

To our knowledge, this is the first study that linkspopulation-based individual-level health survey datawith actual ED utilization and the first Canadian popu-lation-based study on the subject. Several recent U.S.population-level studies of ED utilization found thatthose with worse health status, low income, and unmethealth needs were more likely to use the ED.7,8 Thesame studies found that ED users were no more likelyto be uninsured, nor lack a usual source of care, thannon-ED users, which contradicted prior hospital-basedstudies that found the opposite.3–8 The Canadian healthcare system differs from the United States in that healthinsurance for hospital and physician care is publiclyfunded and universal. In this environment of fewerdirect financial barriers to medical care, we found thatindividuals who are sicker are more likely to use theED, and those who use the ED are also more likely tobe users of other providers of health care such as fam-ily physicians and specialists. Unlike previous studies,however, we found that low education and incomewere associated with increased ED utilization, evenafter controlling for health status and access to primaryand specialist care.

Although our data allowed us to determine the triagecategories and main diagnoses for ED visits, we did notattempt to define ‘‘appropriateness’’ of particular visits,nor whether ED visits could have been managed inother settings. Our aim was to use the triage score anddisposition to define a group of less urgent patientswhose visit to the ED may indicate access barriers toambulatory care and compare their characteristics withthe general, non–ED-using population.

We hypothesized that in Ontario, where a shortageof primary care physicians is recognized across theprovince, less urgent presentations to the ED could bea reflection of immediate access barriers to alternativesites of care such as family doctors or walk-in clin-ics.14,24,25 This hypothesis is supported by single-sitesurveys from Canada that have demonstrated that whileless urgent ED use is not associated with lack of a fam-ily doctor, it is associated with an inability to accesstimely primary care.26,27 A Canadian population-basedstudy demonstrated that recent primary care reformefforts in Ontario have led to increased rostering ofindividuals of higher SES than expected for the popula-tion, also suggesting reduced primary care access forthose with lower SES.28 We further hypothesized thatpeople with low SES would be more likely to experi-ence barriers due to difficulty navigating the health caresystem. The finding that individuals with low education

01020304050607080

1 2 3 4 5Perc

ent w

ith E

D V

isit

(%)

Self-perceived health (1=excellent, 5=poor)

Low educationHigh education

Figure 1. Percentage of males with ED visits versus self-perceived health.

304050607080

th E

D V

isit

(%)


01020304050607080

1 2 3 4 5Perc

ent w

ith E

D V

isit

(%)



01020304050607080

1 2 3 4 5Perc

ent w

ith E

D V

isit

(%)



Figure 2. Percentage of females with ED visits versus self-perceived health.


and income, which we used to represent SES, are morelikely to make ED visits, irrespective of acuity level andhaving a regular doctor, does not support this hypothe-sis. To our knowledge, this is the first time that thisrelationship has been reported in a Canadian popula-tion; however, this concurs with findings from othercountries with public health insurance that report anassociation between SES and overall ED utilization.29

The finding that after-hours and weekend ED visits donot vary by education or income also does not supportan access barrier hypothesis, unless these barriersapply more or less equally to all SES groups.

If access barriers and health status are not a pri-mary cause of increased ED use among people withlow SES in our setting, then what is the explanationfor this finding? It is understood that a greater burdenof disease rests on those who are poorly educated andhave lower income.30 We attempted to control forhealth needs in both primary and sensitivity analyses,but none of these analyses resulted in reduced effectsof education or income. While residual confoundingby need may be an explanation, it is also possible thatperception of urgency and preference for source ofcare vary by SES. These are both concepts that havebeen demonstrated to be associated with less urgentED use in the literature.26,27,31,32 Patients with lowereducation or income may have less knowledge of oraccess to alternatives to EDs, such as tele-health phonelines (e.g., lack of knowledge or telephone) or walk-inclinics (e.g., distance from home) or self-care for self-limited conditions (e.g., over-the-counter medicationsand supplies). Our finding that low-education groupshave higher ED utilization even among those whoreport good or excellent health supports this possibleexplanation. Additionally, measures of health needmay not predict injuries and higher injury rates amongthose with low SES, which may also be contributingto our findings.

Since the inception of EDs as organized and integralsites of care within the health care system, they havebeen recognized for their desirable features.33 Healthcare workers and laboratory and radiology servicesare on-site to provide comprehensive care 24 hours aday, 7 days a week. The accessibility of ED services inCanada and elsewhere may contribute to increased useby low SES groups due to their ‘‘one-stop’’ nature.Visits to EDs may obviate the need for scheduled officevisits or multiple scheduled appointments for furtherinvestigations and ⁄ or consultations. This factor may beimportant for working people with low education orincome due to lack of flexibility to plan ahead, con-straints from inflexible work hours, child care availabil-ity, and ability to organize transportation. Such EDutilization is often considered less desirable from ahealth system perspective; however, low-urgency orlow-complexity patients do not contribute significantlyto overall ED wait times or overcrowding.34,35 In addi-tion, such utilization of EDs may be the most appropri-ate course of action for the individual and may lead tofewer missed work days, lower stress, and lowerincurred costs in the long term. In addition to servicesprovided, there may be more subtle reasons for prefer-ring EDs as sites of care. If EDs are equally welcoming

than office settings to individuals from all social, cul-tural, and economic backgrounds, then low SES groupsmay have a preference for this type of care. Furtherresearch into the contributing factors to ED use isneeded.

LIMITATIONS

The CCHS excluded certain groups and had a degree ofnon-response, all of which could influence the general-izability of these results. Statistics Canada applies bothsampling and poststratification weights to minimizethese potential biases. The independent variables usedin this study are based on self-report, which is suscepti-ble to social desirability, nonresponse, and relatedbiases. For example, just over 10% of respondents didnot complete the question regarding income.

The databases that were used to provide health careutilization data are administrative in nature and werenot designed specifically for research purposes. How-ever, since Ontario’s health system uses a single payer,the use of insurance claims is a nearly complete reflec-tion of actual utilization.

Triage categories in CTAS were designed to directtimely ED care to certain groups. Our use of thesecategories to distinguish more urgent from less urgentvisits has not been externally validated.

CONCLUSIONS

We found that worse health status is the largest predic-tor of ED utilization in Ontario, but low education andincome were independently associated with increasedED visits. The increased use seen in low education andincome groups is evident across acuity levels, suggest-ing that increased health care needs are underpinningthis utilization. This study provides support to findingsin other countries demonstrating that ED users aresicker and socially disadvantaged and are not relyingon EDs simply as a primary care provider. Our findingssupport an international policy direction that should,in the short term, better understand the factors lead-ing to a preference for ED-based care among lowsocioeconomic status patients. In the long-term,improved health promotion, preventive health care and,chronic disease management, coupled with efforts toreduce socioeconomic disparities, may help to alleviatethe greater burden of disease that is carried bydisadvantaged populations.

The authors acknowledge Mohammed Agha for his assistance withthe data sets.

References

1. Canadian Institute for Health Information. Under-standing Emergency Department Wait Times.Ottawa, Ontario: Canadian Institute for HealthInformation, 2005.

2. Institute of Medicine. The Future of EmergencyCare in the United States Health System. AnnEmerg Med. 2006; 48:115–20.


3. Richardson L, Hwang U. Access to care: a review ofthe emergency medicine literature. Acad EmergMed. 2001; 8:1030–6.

4. Grumbach K, Keane D, Bindman A. Primary careand public emergency department overcrowding.Am J Public Health. 1993; 83:372–8.

5. Torrens P. Variation among emergency room popu-lations: a comparison of four hospitals in New YorkCity. Med Care. 1970; 8:60–75.

6. Pane G, Farner M, Salness K. Health care accessproblems of medically indigent emergency depart-ment walk-in patients. Ann Emerg Med. 1991;20:730–3.

7. Weber E, Showstack J, Hunt K, Colby D, CallahamM. Does lack of a usual source of care or healthinsurance increase the likelihood of an emergencydepartment visit? Results of a national population-based study. Ann Emerg Med. 2005; 45:4–12.

8. Zuckerman S, Shen YC. Characteristics of occa-sional and frequent emergency department users:do insurance coverage and access to care matter?Med Care. 2004; 42:176–82.

9. Madore O. The Canada Health Act: Overviewand Options? Available at: http://www.parl.gc.ca/Content/LOP/ResearchPublications/944-e.htm. AccessedMay 23, 2011.

10. HealthCanada. 2003 First Ministers’ Accord onHealth Care Renewal. Available at: http://www.hc-sc.gc.ca/hcs-sss/delivery-prestation/fptcollab/2003accord/index-eng.php. Accessed May 23, 2011.

11. Glazier RH, Tepper J, Agha M, Moineddin R. Pri-mary care in disadvantaged populations. In: Jaakki-mainen L, Upshur R, Schultz S, eds. Primary Carein Ontario. Toronto: Institute for Clinical EvaluativeSciences, 2006.

12. Winkleby M, Jatulis D, Frank E, Fortmann S.Socioeconomic status and health: how education,income, and occupation contribute to risk factorsfor cardiovascular disease. Am J Public Health.1992; 82:816–20.

13. Schoen C, Doty M. Inequities in access to medicalcare in five countries: findings from the 2001Commonwealth Fund International Health PolicySurvey. Health Policy. 2004; 67:309–22.

14. College of Physicians and Surgeons of Ontario.2005 Survey of Ontario’s Physicians: Access Chal-lenges Ahead. Toronto: College of Physicians andSurgeons of Ontario, 2006.

15. Ware J, Snow K, Kosinski M, et al. SF-36 HealthSurvey Manual and Interpretation Guide. Boston,MA: New England Medical Center, 1993.

16. McHomey C, Ware J, Raczek A. The MOS 36-itemshort-form health status survey (SF-36) II: psycho-metric and clinical tests of validity in measuringphysical and mental health constructs. Med Care.1993; 31:247–63.

17. Grafstein E, Innes G, Westman J, Christenson J,Thorne A. Inter-rater reliability of a computerizedpresenting-complaint-linked triage system in anurban emergency department. Can J Emerg Med.2003; 5:323–9.

18. Statistics Canada. Canadian Community HealthSurvey (CCHS): Cycle 1.1: extending the wealth

of health data in Canada. Available at: http://www.statcan.ca/english/concepts/health/cchsinfo.htm.Accessed May 23, 2011.

19. World Health Organization. International StatisticalClassification of Diseases and Related Health Prob-lems: Tenth Revision, 2nd edn. Geneva, Switzerland:World Health Organization, 2004.

20. The Johns Hopkins University. Johns Hopkins ACGCase-Mix Adjustment System. Available at: http://www.acg.jhsph.edu. Accessed May 23, 2011.

21. Reid R, MacWilliam L, Verhulst L, Roos N, AtkinsonM. Performance of the ACG case-mix system in twoCanadian provinces. Med Care. 2001; 39:86–99.

22. Starfield B, Weiner J, Murla P. Ambulatory caregroups: a categorization of diagnoses for researchand management. Health Serv Res. 1991; 26:53–74.

23. Weiner J, Starfield B, Steinwachs D, Abramson J.Development and application of a population-ori-ented measure of ambulatory care case-mix. MedCare. 1991; 29:452–72.

24. Ontario College of Family Physicians. Where HaveAll the Family Doctors Gone? Toronto, Ontario:Ontario College of Family Physicians, 2001.

25. Schultz SE, Tepper J, Guttman A, Jaakkimainen L.Characteristics of primary care practice. In: Jaakki-mainen L, Upshur R, Klein-Geltink JE, et al., eds.Primary Care in Ontario: ICES Atlas. Toronto:Institute for Clinical Evaluative Sciences, 2006.

26. Steele S, Anstett D, Milne WK. Rural emergencydepartment use by CTAS IV and V patients. Can JEmerg Med. 2008; 10:209–14.

27. Field S, Lantz A. Emergency department use byCTAS Levels IV and V patients. Can J Emerg Med.2006; 8:317–22.

28. Glazier RH, Redelmeier DA. Building the patient-centered medical home in Ontario. JAMA. 2010;303:2186–7.

29. Shah SM, Cook DG. Socio-economic determinantsof casualty and NHS Direct use. J Public Health(Oxf). 2008; 30:75–81.

30. Marmot M, Smith G, Stansfeld S, et al. Healthinequalities among British civil servants: the White-hall II Study. Lancet. 1991; 337:1387–93.

31. Tsai JC, Liang YW, Pearson WS. Utilization ofemergency department in patients with non-urgentmedical problems: patient preference and emergencydepartment convenience. J Formos Med Assoc.2010; 109:533–42.

32. Backman AS, Blomqvist P, Lagerlund M, Carlsson-Holm E, Adami J. Characteristics of non-urgentpatients. Cross-sectional study of emergencydepartment and primary care patients. Scand JPrim Health Care. 2008; 26:181–7.

33. Pollock D. Barriers to health care access: whatcounts and who’s counting? Acad Emerg Med.2001; 8:1016–8.

34. Schull M, Kiss A, Szalai J. The effect of low-com-plexity patients on emergency department waitingtimes. Ann Emerg Med. 2007; 49:257–64.

35. Asplin B, Magid D. If you want to fix crowding,start by fixing your hospital. Ann Emerg Med. 2007;49:273–4.


Documents

A Population-based Study of the Association Between Socioeconomic Status and Emergency Department Utilization in Ontario, Canada