Factors associated with non-urgent utilization of Accidentand Emergency services: a case-control study in Hong

Kong

Albert Leea,*, Fei-Lung Laub, Clarke B. Hazlett c, Chak-Wah Kamd,Patrick Wonge, Tai-Wai Wongf, Susan Chowa

aDepartment of Community and Family Medicine, The Chinese University of Hong Kong, Hong KongbAccident & Emergency Department, United Christian Hospital, Hong KongcFaculty of Medicine, The Chinese University of Hong Kong, Hong KongdAccident & Emergency Department, Tuen Mun Hospital, Hong Kong

eAccident & Emergency Department, Yan Chai Hospital, Hong Kong, Hong KongfAccident & Emergency Department, Pamela Youde Nethersole Eastern Hospital, Hong Kong

Abstract

Accident and Emergency Departments (A&E) have been a popular source of primary care, and studies haveshown that up to two thirds of patients attending A&E have problems that could be managed by generalpractitioners (GPs). Although many studies have found that patients of lower socio-economic class with less socialsupport have a higher utilization rate of A&E, some recent studies have revealed contrary evidence. In this study2410 patients were randomly selected from four A&E at di�erent times. The gold standard in di�erentiating trueemergency cases and GP cases was based on a retrospective record review conducted independently by a panel ofemergency physicians. Two emergency physicians reviewed each case independently, and if their independent ratingswere in agreement, this became the gold standard. Patients classi®ed as GP cases were given a telephone interview,and a sample was selected and matched with cases from general out patient clinics (GOPC) in the public sector bymorbidity. Reasons for not attending a private GP included closure of clinic, deterioration of symptoms, GPs'inability to diagnose properly, and patients' wish to continue medical treatment in the same hospital. Reasons whynon-urgent patients did not choose to attend the nearby public GOPC included a�ordability, closure of the GOPC,patients' wish to continue treatment at the same hospital, GOPC too far away, no improvement shown after visitsto GOPC doctors, and GOPC doctors' inability to make proper diagnoses. The reasons for high level of utilizationof A&E services are complex and re¯ect problems of delivery of GP services. There is an urgent need for GPs to setup a network system to provide out of hours services, and also for a better interfacing between primary andsecondary care, and between public and private sectors, so that patients can be referred back to GPs. Interimclinical services provided to those non-urgent cases by nursing practitioners or by GPs working in A&E could alsofacilitate discharge of patients to primary care facilities. 7 2000 Elsevier Science Ltd. All rights reserved.

Keywords: Emergency services; Utilization of services; Hong Kong

Social Science & Medicine 51 (2000) 1075±1085

0277-9536/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved.

PII: S0277-9536(00 )00039-3

www.elsevier.com/locate/socscimed

* Corresponding author. Tel.: +852-2693-3708; fax: +852-

2694-0004.

E-mail address: alee@cuhk.edu.hk (A. Lee).

Introduction

The hospital Accident and Emergency Department(A&E) is meant to serve patients with immediate lifethreatening or critical conditions. Non-urgent utiliz-

ation refers to inappropriate attendance at A&E bypatients whose conditions are neither accidents noremergencies and often require no speci®c hospital

treatment. The signi®cant increase of inappropriateA&E attendance is considered to be a serious threat tothe health care system. Some studies have found that

up to two thirds of patients who attend A&E haveproblems that could be managed appropriately by gen-eral practitioners (Myers, 1982; Green & Dale, 1992;Driscoll, Vincent & Wilkinson, 1987; Bowling, Isaacs,

Armstorn, Roberts & Elliott, 1987a; Bowling, Isaacs,Roberts & Elliott, 1987b; Bentzen, Christiansen & Ped-ersen, 1987; Andersen & Gaudry, 1984; Dale, 1992;

Cohen, 1987; Davies, 1986; National Center for HealthStatistics (McGraig), 1994; American College of Emer-gency Physicians, 1990).

In Shatin, Hong Kong (a satellite town with popu-lation of 500,000), a study has shown that over 20%of A&E consultations sought by elderly patients were

not urgently needed (CUHK, 1994). Another localstudy illustrated that 50% of A&E attendees werenon-trauma conditions (Leicester et al., 1991). Despitean education e�ort by the Hospital Authority, which

oversees all the public hospitals in Hong Kong, thenumber of A&E attendees has escalated in recentyears. Unlike some other countries, the triage of

patients from emergency departments to primary careis considered inappropriate and impracticable in thepresent Hong Kong situation. Therefore, the authors

have conducted a study to determine the level of inap-propriate use, the nature of the morbidity pattern, thevalidity of the nurse triage system and the reasons whyalternate and appropriate primary care services are not

being accessed within this local context. This paperwill report the factors associated with inappropriateutilization of A&E services in Hong Kong.

Literature review

Many studies have demonstrated that the emergencyroom is disproportionately used by low income

patients, who very often do not have a family doctoror are not covered by medical insurance (Leicester etal., 1991; Padgett & Brodsky, 1992; Hull, Jones Rees

& Moser, 1997; Beland, Lemay & Boucher, 1998).Aside from these social factors, cultural and psychoso-cial reasons also play a role in the non-urgent utiliz-

ation of the A&E (Padgett & Brodsky, 1992; Haddy,Schmaler & Epting, 1987). Another study has demon-strated that patients who have no previous experience

of their symptoms may often misperceive their con-dition as an emergency and may think that A&E care

is more appropriate for their condition (Klijakovic,Allan & Reinker, 1981).Members of the general public often see the A&E as

the most suitable facility to get medical aid quickly.Other than these factors, the perception of distance tothe hospital and the GP was also an important predic-

tor of patients' choice of place for medical consultation(Bowling et al., 1987; Beland, Philibert, Thouez &Maheux, 1990). However, one study conducted in

Kuwait found that higher educated and more a�uentpatients were more likely to utilize A&E services fornon-urgent purposes than those from lower socio-econ-omic classes (Shah, Shan & Behbehani, 1996). A study

in New Zealand reported that the cost of visiting thegeneral practitioner did not appear to in¯uence parentsuse of the A&E. Families of low socio-economic status

were more likely to see the GP ®rst (Klijakovic et al.,1981). Social support has been shown to have an im-portant e�ect on the way a family uses medical ser-

vices. However, con¯icting evidence has been reportedon the use of emergency services where grandmothersare involved in child care. In an American study, there

was a greater use of emergency services, but the reversewas found in a British study (Fergusson, Li & Taylor,1998).None of the above studies used case control from

the primary care setting to compare the variables onnon-urgent utilization. The King's College HospitalA&E project in the UK investigated the reasons why

patients attended the A&E with primary care pro-blems. The study compared the non-urgent A&Epatients with those who presented with new problems

in general practice (Dale, Green, Glucksman & Higgs,1991). This comparison could not represent the usualGP problems, because new problems tended to be oflonger duration and were less likely to be injuries in

the primary care settings. More new factors wouldhave emerged from the study if matching by complainthad been undertaken.

Organization of Hong Kong health care system

Hong Kong runs a dual system, private and publicfor both primary and secondary care. The govern-

ment's basic mission statement on health care has beenthat no one should be deprived of care because of lackof means. Patients in public hospitals or clinics only

pay a small nominal fee (less than US$8 per day) forconsultation, including medication, investigations, pro-cedures and other overhead charges, e.g. in-patient

meals. This fee can also be waived if social assistanceis needed, and the A&E services are free.The Hospital Authority is a statutory body, indepen-

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±10851076

dent from the civil service, which oversees the manage-ment and operation of all public funded hospitals. The

Government is responsible for funding the public hos-pitals. The public sector provides over 90% of the hos-pital care for the population. However, the picture is

reversed for primary medical care with private generalpractitioners providing 70% of primary medical care(Lee, 1997). The Department of Health is a govern-

ment department responsible for public health issuesand primary health care services in the public sector.The Department of Health provides 15±20% of pri-

mary medical care to the population, and others utilizealternative medicine for primary medical care (Lee,1997). Hong Kong has enjoyed very good economicgrowth with very low taxation and a gross domestic

product (GDP) per capita that is very close to that ofthe United States, and above that of Canada and Uni-ted Kingdom. Spending on health is about 3.9% of

GDP (1990±91), less than Britain and the OECD aver-age (Hay, 1992).

Methodology

In this study, the study population was a cross sec-tion of patients attending four hospitals in di�erentgeographical locations in Hong Kong. The accessible

population was comprised of patients attending theA&E departments located in three geographicalregions in Hong Kong: Hong Kong Island with a

population of 1.3 million, Kowloon with a populationof 1.9 million, and New Territories with a populationof 2.9 million. The hospitals were the Pamela Youde

Nethersole Eastern Hospital serving the population ineastern side of the Hong Kong Island, United Chris-tian Hospital serving the east Kowloon region, TuenMun Hospital and Yan Chai Hospital serving the

population in the New Territories.Patients were selected randomly from these four

A&E departments in Hong Kong. The gold standard

establishing the true urgency rate was based on a blindretrospective record review conducted independentlyby a panel of emergency physicians. Apart from taking

into account the skills expected of emergency phys-icians, the di�erentiation of true emergency and GPcases was also based on the Handbook on VocationalTraining in Family Medicine by the Hong Kong Col-

lege of Family Physicians (HKCFP). The handbookhas a thorough description of the necessary knowledgeand skills family physicians require in order to com-

plete vocational training in family medicine. Whenpanel members' independent ratings were in agreement,the classi®cation was regarded as the Gold Standard

for utilization. If no agreement was reached, the princi-pal investigator (an academic family physician with ex-perience in A&E) or the trained research sta� would

classify the cases according to the Handbook from the

HKCFP.A sample size of 2410 patients was necessary in

order to establish the level of acceptance error at

20.02 in the total sample [N=(z/e )2(p)(1ÿp)=(1.96/0.2)2 p = 0.05, using the unit normal deviate (z ) of

1.96 corresponding to a 95% level of con®dence andthe most conservative rate ( p ) at 0.50]. Therefore,241 h of data collection were required at ten patients

per hour. The number of attendees varied hourly fromone day to another. In order to avoid sample bias, itwas important to obtain a representative sample that

took into account these variations. The averages forthe low, medium and high utilization period were esti-

mated to be twelve, twenty-®ve and forty visits perhour respectively. On average there are about 600attendees per day within the four study hospitals. In

order to obtain a representative sample, the study pro-portionally sampled the number of hours selectedwithin each utilization period. Therefore, 39, 75 and

127 h from low, medium and high utilization periodrespectively were selected.

This procedure produced sample sizes of 390, 750and 1270 subjects from the low, average and high util-ization periods, which were determined by the z score

of the individual hour. The utilization pattern of thefour di�erent hospitals was tabulated hourly for eachday of the week for the most recent month. To deter-

mine the grand mean of the tabulated hour for eachday, the recent available one year record of A&E util-

ization was tabulated hourly for the year. Using thegrand mean and standard deviation per hospital foreach week, the z score for each 1-h block within the

week was calculated. The four z scores for the sameday and hour in each week were averaged across thefour weeks. A low utilization period had a z score of

less than 1 standard deviation, while a z score ofgreater than 1 standard deviation represented a high

utilization period. After the utilization periods hadbeen de®ned for each hospital, 1-h blocks were ran-domly selected by the computer in order to obtain a

representative sample of utilization pattern for each in-dividual hospital. During the selected hour of thestudy period, the ®rst patient presenting for regis-

tration was approached for interview. The next inter-view followed immediately. Patients were excluded if

they were unable to answer questions and had noguardian present who could answer for them, or ifthey refused to be interviewed.

Written informed consent was gained from eachpatient and, in addition, permission for a follow up tel-

ephone interview within 2±4 weeks. During the tele-phone interview, the reasons for utilizing the A&Eservices in preference to private GPs or public GOPC

were discussed, and related questions on biographicand demographic characteristics, such as age, sex, edu-

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±1085 1077

cation level, occupation, insurance cover, having a

family doctor or not, location of residence were asked.Details of the methodology are described elsewhere(Lee et al., 1999).

Of the 2410 patients in the study, 1378 (57%) wereclassi®ed as GP cases (Lee et al., 1999). In order to

establish statistically signi®cant factors distinguishingthe non-urgent patients from the patients with similarmorbidity who choose to visit an outpatient facility,

the study required at least 700 patients from eachgroup, assuming a small e�ect size of 15% when stat-istical power of 0.80 is desired (using a 0.05, 2-tailed

level of signi®cance).Subsequent to matching for similar morbidity pat-

terns to A&E patients, 726 patients were selected fromthe General Outpatient Clinic (GOPC) at two generalhospitals. Their diagnoses were coded according to the

International Code Primary Care (ICPC). Out of the1322 GP cases that met the criteria for telephone con-tact by researchers, 726 patients were randomly

selected proportionally from the fourteen system cat-egories. For example, if 30% of the 1322 non-urgent

patients had problems related to the respiratory sys-tem, of the 726 patients selected 30% also had respir-atory problems.

The research sta� carried out telephone interviewswith both the A&E GP cases and the matched GOPC

groups. The same questionnaire was given to bothgroups of patients, except that the GOPC matchedcases were asked why they did not utilize A&E ser-

vices. This enabled analysis of the di�erence in thereasons for utilization of services, and also of thedi�erence in biographic and demographic character-

istics between the two groups.Using Statistical Package for Social Science (SPSS)

software, the age and sex distribution of the studypopulation and comparison of 1997 A&E attendeeswere ascertained. In addition, the relationship between

sex and age and level of utilization, relationshipbetween time period and age, detailed studies of theGP cases of A&E and true A&E patients in relation to

sex, age, utilization period and time period were con-ducted. The 726 A&E GP cases selected to match with

the GOPC cases with same morbidity were used for acontrol study. A chi square test was used to determinethe statistical signi®cance of reasons for A&E GP

cases use of A&E services in preference to nearby GPor GOPC services. Besides the various reasons forattending the di�erent facilities, other demographic

characteristics, such as educational background andsocial economic status of the two groups of patients,

were analyzed.Multiple Logistic Regression was used to distinguish

the di�erence between GP cases and matched GOPC

primary care patients on signi®cance and odds ratio ofthe variables. The variables for analysis were: patients'

rating of the facility service, socio-economic status,education level, age group of patients, whether the

patients were health insured or had a family doctor,and reasons for utilization (reasons for not attendinglocal GP and reasons for not utilizing the nearby

GOPC or A&E services).All of the investigators (experienced A&E or family

physicians) contributed to the preparation of the data

collection instruments, so as to ensure the face validityof both the interview and the study protocol. Severalfocus groups conducted at the A&E departments in

the four participating hospitals also further validatedthe contents of the questionnaires.

Results

The research sta� identi®ed 2892 patients according

to the protocol at the four A&E departments to con-sent to interview and telephone follow up if required.

Fig. 1. (a) The 1997 A&E attendees in Hong Kong compared

with the study population (HAHO 1997 statistics). (b) The

1997 Male A&E attendees in Hong Kong compared with the

study population (HAHO 1997 statistics). (c) The 1997

Female A&E attendees in Hong Kong compared with the

study population (HAHO 1997 statistics).

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±10851078

Of these 2410 were recruited across the di�erent utiliz-ation period with response rate over 80%. The reasons

for refusal included lack of time, unwilling to be dis-turbed and personal reasons. The patients sampled forthe study were similar in pro®le to the 1997 A&E

attendees in Hong Kong in the sex and age distri-bution (Fig. 1a±c).

Table 1 shows the proportion of the A&E and non-urgent cases in the study population (43% vs 57%). A

higher proportion of males attended the A&E thanfemales across most of the age groups, except for over65 years of age, but a higher proportion of females

attended the A&E for cases that could be treated byGPs (58.4% vs 41.6%, p = 0.2). A higher proportion

of patients in age groups 0±9 and 10±19 utilized A&Efor conditions that could be treated by GP. The trend

was signi®cantly reversed for those over the age of 45(Table 2). Table 3 shows that patients who lived within5 km of the hospitals used the A&E services more for

general practice purposes, whereas those patients livingmore than 6 km from the hospitals tended to use the

services more for emergency purposes. Fig. 2a and bshow the education and employment status of the non-

urgent patients attending the A&E. Three quarters ofthese patients had secondary school or above edu-cational background and the majority of them (62.4%)

were full time workers. 50.8% resided in private ac-commodations. These results do not support the view

that patients from lower socio-economic class utilizeA&E more frequently for GP purposes.

Amongst the GP cases attending the A&E, 726 caseswere chosen to be matched by morbidity with patientsattending the general outpatient clinics in hospitals

under the Hospital Authority during the 3-month sche-

dule (Fig. 3). When the socio-economic status of theselected non-urgent A&E and GOPC patients was

compared, there was a signi®cantly higher proportionof the patients (26.0% vs 22.9%, p < 0.01) living inprivate accommodation utilizing the A&E for non-

urgent purposes. Among all GP cases, 39.1% claimedto have a family doctor, while only 17.2% patients

attending GOPC claimed to have a family doctor ( p<0.01). A higher proportion of more educated patients(74.9% vs 51%, p < 0.01) utilized the A&E services

for GP purposes than those utilizing GOPC. Patientswith skilled jobs utilized the A&E for GP purposes

more than those in GOPC (35.3% vs 15%, p < 0.01).The majority of patients (87.7%) expressed a wish to

be referred back to their own doctors once their con-dition had been stabilized.

For those patients who attended A&E with con-ditions that could be treated by GPs (726 selectedpatients in comparison with the 726 GOPC patients),

the main (and statistically signi®cant, p<0.05) reasonsfor utilizing the A&E service were: perceived emer-

gency status of their disease; feeling sick on publicholidays or at night; living in close proximity to the

hospitals; availability of proper diagnosis and e�cientservice at the time of day it was needed (Fig. 4). Otherfactors which also demonstrated statistical signi®cance

were the desperate need for help, the feeling that thesituations could best be handled in the A&E facility,

and the fact that patients had been sent to the depart-ment from school or from their work place.

When patients were asked for reasons for notattending a private GP, closure of the clinic was themain reason given. In addition, patients mentioned de-

terioration of their symptoms, the GPs' inability to

Table 1

Relationship between sex and utilization of AED for cases

that could be treated by GPs/true A&E purposea

Type of cases/sex Male Female Total

A&E cases 581 (44.1) 455 (41.6) 1036 (43.0)

GP cases 735 (55.9) 639 (58.4) 1374 (57.0)

Total 1316 (100) 1094 (100) 2410 (100)

a p=0.2.

Table 2

Relationship between age and utilization of AED for cases that could be treated by GPs/true A&E purposea

Age/Types of cases 0±9 10±19 20±44 45±64 Above 65 Total

A&E cases 176 (31.4) 81 (31.4) 318 (37.1) 206 (52.3) 255 (75.0) 1036 (43.0)

GP cases 384 (68.6) 177 (68.6) 540 (62.9) 188 (47.7) 85 (25.0) 1374 (57.0)

Total 560 (100) 258 (100) 858 (100) 394 (100) 340 (100) 2410 (100)

a p<0.01.

Table 3

Relation of level of utilization and distance away from hospi-

tala

Types of cases/Distance from AEDa < 5 km >6 km

A&E cases 868 (41.1) 190 (53.8)

GP cases 1211 (58.9) 163 (46.2)

Total 2057 (100) 353 (100)

a p<0.01.

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±1085 1079

diagnose properly and the fact that they wished tocontinue medical treatment in the same hospital, whererecords of previous treatment were kept. A�ordability

was the most pronounced reason for the GOPCpatients, but did not apply to A&E patients (Fig. 5).The main reason why non-urgent patients did not

choose to attend the nearby GOPC was closure of theGOPC at night and in the public holidays. Otherreasons showing statistical signi®cance were: patientswanting to continue treatment at the same hospital,

GOPC was too far away, no improvement shown afterbeing seen by the GOPC doctors and the unavailabilityof proper diagnosis. For GOPC patients, on the other

hand, an extremely pronounced reason for not choos-ing to use A&E services was that that facility did notsuit their needs (Fig. 6).

Multiple logistic regression was performed to ana-lyze which factors were independently associated withutilizing A&E for general practice conditions, using

patients with similar morbidity from the GOPCs ascontrol. Table 4 summarizes the signi®cant associatedfactors. The results indicate that the odds ratio washigher for children aged 0±9 (OR 5.44, 95%CI 1.6±

18.2) and fewer part time workers utilized A&E ser-vices than full time workers (OR 0.38, 95%CI 0.19±

0.78). Perceived emergency status of the disease had avery high odds ratio. Several factors are statistically

signi®cant in the utilization of A&E over that ofGOPCs. These factors include: (1) greater e�ciency ofdiagnoses; (2) self perceived severity of the conditions

to be managed by A&E; (3) closure of GP clinics; and(4) patients being desperate for help. On the cost side,the factor of low cost is associated with lower odds for

the GP cases utilizing A&E (OR 0.08 95% CI 0.03±0.17).

Discussion

Although the four A&E departments were not ran-domly selected, they are located in di�erent parts ofHong Kong representing both urban and rural areas.

The age and sex distribution of the study populationwas similar to that of the total of A&E attendees. Thetime blocks were randomly selected to represent pro-

portionally the high, medium and low utilizationperiods. The response rate was over 80%. Thereforethe study population was a reasonable representativesample of the target population.

A unique feature of this study was the use ofpatients with similar morbidity attending the GOPC ascontrol. This enabled a more valid analysis of factors

associated with utilization of A&E for general practicepurposes. An idea could have been to match each casewith a control. However, not only did the time scale of

the study prevent us from doing this, but also we feltthat overmatching might lead to problems in discrimi-nating the di�erence between the cases and control.

More matching controls involving private GPs'patients introduces di�erential in payment as a majorconfounding factor. In a dual health care system, it isimportant to examine this confounding factor.

In contrast to many overseas studies, results of thisstudy reveal that those patients who utilized A&E forGP problems were not in the lower socio-economic

class and a higher proportion had their own familydoctors when compared with GOPC patients (Green &Dale, 1992; Andersen & Gaudry, 1984; Haddy et al.,

1987). Figs. 2±6 show that patients with non-urgentproblems utilized A&E because general practice ser-vices were not accessible, especially during the timewhen symptoms worsened. Factors such as conven-

ience, shorter waiting time and e�ciency of serviceswere also main reasons for utilizing A&E. A multivari-ate analysis demonstrates perceived urgency, closure of

clinics, desperate need for help and conditions besthandled by A&E, as independent associating factors.However, low cost is signi®cantly associated with

lower odds of utilizing A&E for GP purposes. Thisimplies that, to save money, patients tend to chooseGOPC rather than attending A&E.

Fig. 2. (a) Education level of GP cases. (b) Employment sta-

tus of GP cases.

Fig. 3. Comparison between non-urgent cases in A&E and

GOPC cases.

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±10851080

A higher proportion of those utilizing A&E for GP

purposes had a family doctor than those attending

GOPC, and their odds of utilizing A&E were in fact

higher. This further indicates utilization for conven-

ience. Although it was reported that over 80% of

patients utilizing A&E for GP purposes wanted to be

referred back to their family doctors after their con-

ditions became stable, this group tended to utilize

A&E more because the general practice services were

not available when needed. About 60% of the subjects

rated the primary care provision in Hong Kong as

only fair.

The study also demonstrates that a substantial pro-

portion of the higher socio-economic group and young

adults utilized A&E services for non-urgent purposes

rather than the socially disadvantaged. Hong Kong is

a well-established ®nancial center with a robust econ-

omy, so private GP services are not out of reach for

most persons. The GOPC provides a safety net for the

disadvantaged. The problem of inappropriate utiliz-

ation of A&E is a problem of delivery of comprehen-

sive primary health services, rather than of

a�ordability. There is a need to develop the means for

patients to gain access to GP services after surgery

hours. Extending GOPC working hours may not solve

the problem entirely, as this only helps the lower

socio-economic patient. A�uent patients will still use

A&E services rather than GOPC if their family doctors

are not available.

Although living within 5 km was found to be associ-

ated with a higher proportion of patients utilizing

A&E for GP purposes in univariate analysis, it was

Fig. 4. A comparative study of the non-urgent A&E & GOPC patients on reasons for utilizing the AED & GOPC service.

Fig. 5. A comparative study of the non-urgent A&E and GOPC patients on reasons for not attending private GP.

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±1085 1081

not found to be statistically signi®cant in multivariate

analysis. Hong Kong is not a very large city and health

care facilities are fairly ubiquitous. Factors other than

proximity to health care setting are usually associated

with inappropriate utilization of A&E.

As 70% of primary medical care is provided by pri-

vate GPs (Lee, 1997) and most of them are in solo

practice, they cannot be expected to provide a 24-h ser-

vice. However, they could form a network to provide

on-call system by rotation. With advanced technology

and consent of patients, the network could link to the

medical record system to facilitate decision making by

on-call doctors. Alternatively GPs could establish

record linkage with nearby private hospitals so that

resident medical o�cers could have access to patient

medical histories, and continuing care could be pro-

vided. Residents could contact the patient's family

doctor to exchange information. The family doctor

could then take over the management the following

morning with back up from residents outside o�ce

hours. This kind of arrangement needs to be well co-

ordinated to enable greater care in the community and

to decrease the utilization of A&E except for `true'

emergencies for those who can a�ord to see private

GPs.

A valid `screening' examination is needed to identify

those patients that are `truly' non-urgent. Nurse triage

was shown in this study to be very e�ective in di�eren-

tiating semi-urgent and non-urgent cases (Lee et al.,

1998). If through the triage system a group of patients

can be selectively de¯ected from A&E without signi®-

cant adverse outcomes, this may o�er one solution to

the problem of overcrowding (Grumbach, Keane &

Bindman, 1993; Derlet, Nishio, Cole, & Silva, 1992).

In the United States, federal legislation requires that a

screening examination should be done when patients

present themselves to A&E, but the federal law does

not require treatment to be rendered unless the patient

has an emergency condition (Frew, 1991). Therefore a

highly sensitive screening examination should o�er a

Fig. 6. A comparative study of the non-urgent A&E & GOPC patients for not utilizing the nearby GOPC or A&E service.

Table 4a

Factors Odds ratio 95% C.I.

Employment Full time (Reference) 1

Part time or housewife 0.38 0.19±0.78

Age 65 or above (Reference) 1

0±9 (1) 5.44 1.6±188.2

Job Unskilled vs skilled 0.4 016±096

Reasons for utilization Perceived emergency 557.6 50.8±6112

Low cost 0.08 0.03±017

E�cient diagnosis 66. 2.7±15.9

Desperate need help 8.3 1.6±41.7

Easy refer to specialist 0.23 0.07±0.69

Best handled in AED 3.23 1.07±9.78

GP clinic closed 9.1 1.96±42.2

a p<0.05.

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±10851082

solution to the overcrowding of A&E without incur-

ring legal and ethical dilemmas. Further study focusedon nurse triage as a screening examination for non-urgent patients should be supported. Also those

screened as non-urgent cases should be charged a gen-eral practice consultation fee at market price, as thereis strong evidence that co-payment has a signi®cant

e�ect on patient behavior (Murphy, Plunkett, Bury,Lynam, Smith & Gibney, 1987; Selby, Fireman &

Swain, 1996; O'Grady, Manning, Newhouse & Brooke,1985).In addition to a focused screening examination and

the special training of triage nurses, other factors areessential if A&E departments are to be relieved of

non-urgent patients. There must be community sup-port through the provision of clinics where referrednon-urgent patients can receive timely care. A&E

departments must have good communication with pri-vate GPs to allow for coordination of referral back ofGP cases with case summaries. It is particularly im-

portant to enhance the availability and coordination ofprimary care services, especially for people on low

income who cannot a�ord private GPs. The range ofdiagnostic facilities and medications should beimproved in GOPC, so that patients will not need to

utilize A&E services for continuing treatment and diag-nosis. Re-distribution of resources is needed, but thiswill improve the e�ciency of the health care delivery

system by developing a better interface between pri-mary health care and the hospital sector. One way of

achieving appropriate interfacing is the establishmentof an integrated clinical information system betweenlocal GPs and public hospitals.

As well as performing triage, some nurses should betrained to become nursing practitioners with the ca-pacity to assess and provide basic treatment to non-

urgent patients according to predetermined criteria andto refer patients directly to appropriate clinics and

wards (Smith, 1995; Morris, Head & Holkar, 1989;Vayda, & Gent, 1973). Nursing practitioners would beable to refer patients to other primary care facilities

and provide safe and e�ective service (Brown &Grimes, 1995). Further studies should be conducted toevaluate the outcomes of those patients being managed

by nursing practitioners. If this care is proven to bee�ective and safe, the increased use of nursing prac-

titioners may not only solve the problem of overcrowd-ing of the hospitals, but also improve e�ciency.Another possibility is to employ quali®ed GPs to

work in A&E on a sessional basis. It has been shownin a study conducted by Dale et al. (1995) that employ-

ing GPs in A&E to manage patients' primary careneeds resulted in a reduced rate of investigations, pre-scriptions and referrals. A related study showed that

primary care patients could be managed in this way atreduced cost and with no detrimental e�ect on out-

comes (Dale, Lang, Roberts, Green & Glucksman,1996). General practitioners tend to use time as a diag-

nostic tool and are more experienced in diagnosing`non-disease' and in diagnosing common problems inthe early stages without obvious symptoms and signs.

Employing GPs to work in A&E on a part-time basiscan help GPs keep abreast of developments in hospitalservices and develop closer links with the hospitals.

A&E sta� working closely with GPs may also lead toincreased appreciation of the skills and expertise ofGPs. The A&E sta� may gain greater awareness about

primary care services and this will facilitate shared careand greater integration between hospital and generalpractice, so that patients may be discharged morequickly.

Although patients may be educated to recognize theseverity of their illnesses and to utilize A&E appropri-ately, valid patient self screening is a possibility only in

the distant future. The urgent issue is greater inte-gration of primary and secondary services, so thatpatients can be discharged back to their own family

doctors for continuing treatment, rather than to hospi-tals and especially to A&E. It has been shown thatreferral to primary care providers can be acceptable to

patients (Hansagi, 1990). The provision of interimclinical services to non-urgent patients by nursing prac-titioners or GPs working in A&E will facilitate dis-charge of patients to primary care facilities and send a

clearer signal to patients that A&E is the place for`true' emergencies.

Conclusion

The reasons why patients utilize A&E services fornon-urgent conditions are complex. The higher utiliz-

ation rate by higher socio-economic classes suggeststhat the ®nancial factor is not the main reason forinappropriate utilization. This is further supported by

evidence that low cost is associated with lower oddsratio of utilizing A&E services for primary care pur-poses. The organization of the health care delivery sys-

tem plays a signi®cant role. Limitations in theavailability and accessibility of comprehensive primaryhealth care services cause patients to utilize emergencyservices as an alternative to primary health care ser-

vices. The problem will only be solved if an integratedinfrastructure is established which provides an appro-priate interface between primary and secondary care,

public and private sectors, and also medical and alliedhealth professionals. Nursing practitioners should betrained to transfer appropriate cases to GPs through

the triage process. They should also be trained to man-age some minor problems with back up support fromemergency specialists. More emphasis upon coordi-

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±1085 1083

nation of the di�erent services will enable patients toget maximum bene®t without duplication of resources.

Acknowledgements

The authors would like to thank the sta� of theA&E Departments of United Christian Hospital,

Pamela Youde Nethersole Eastern Hospital, Yan ChaiHospital and Tuen Mun Hospital for their kind co-op-eration and assistance in data collection. We also wish

to thank Health Services Research Fund for theresearch grant for this study.

References

American College of Emergency Physicians (1990). Hospital

and emergency department overcrowding. Annals of

Emergency Medicine, 19, 336.

Andersen, N., & Gaudry, P. L. (1984). Patients attending an

Accident and Emergency Department for primary Medical

Care. Family Practice, 1(2), 79±85.

Beland, F., Lemay, A., & Boucher, M. (1998). Pattern of vis-

its to hospital based emergency rooms. Social Science &

Medicine, 47(2), 165±179.

Beland, F., Philibert, L., Thouez, J., & Maheux, B. (1990).

Socio-spatial perspectives on the utilization of emergency

hospital services in two urban territories in Quebec. Social

Science & Medicine, 30, 53±66.

Bentzen, N., Christiansen, T., & Pedersen, K. M. (1987).

Choice of care for minor trauma: Hospital or general prac-

tice? Family Practice, 4(2), 91±96.

Bowling, A., Isaacs, D., Roberts, J. E., & Elliott, E. J. (1987).

Patient use of a Paediatric Hospital Casualty in the East

End of London. Family Practice, 4(2), 85±95.

Bowling, A., Issacs, D., Roberts, J. E., & Elliott, E. J.

(1987b). Patient use of a Paediatric Hospital Casualty in

the East End of London. Family Practice, 4(2), 85±90.

Brown, S. A., & Grimes, D. E. (1995). A meta-analysis of

nurse practitioners and nurse midwives in primary care.

Nursing Research, 44, 332±339.

Cohen, J. (1987). Accident and emergency services and gen-

eral practice Ð con¯ict or co-operation? Family Practice,

4, 81±83.

CUHK, 1994. Chinese University of Hong Kong Community

and Family Medicine Group Project. Module II Group A.

Accident and Emergency Consultation by the Elderly in

Shatin: Utilisation of Accident and Emergency services.

Dale, J. (1992). Primary care: the old bugbear of accident and

emergency services. British Journal of General Practice, 42,

90±91.

Dale, J., Green, J., Glucksman, E., Higgs, R., 1991. Providing

for Primary Care: Progress in A&E, London. Department

of General Practice and Primary Care, King's College

School of Medicine and Dentistry.

Dale, J., Green, J., Reid, F., Glucksman, E., & Higgs, R.

(1995). Primary care in the accident and emergency depart-

ment: II. Comparison of general practitioners and hospital

doctors. British Medical Journal, 311, 427±430.

Dale, J., Lang, H., Roberts, A. J., Green, J., & Glucksman,

E. (1996). Cost e�ectiveness of treating primary care

patients in accident and emergency: a comparison between

general practitioner, senior house o�cers and registrars.

British Medical Journal, 312, 1340±1344.

Davies, T. (1986). Accident department or general practice.

British Medical Journal, 292, 241±244.

Derlet, R. W., Nishio, D., Cole, L. M., & Silva, J. (1992).

Triage of patients out of the emergency department: three-

year experience. American Journal of Emergency Medicine,

10(3), 195±199.

Driscoll, P. A., Vincent, C. A., & Wilkinson, M. (1987). The

use of the accident and emergency department. Archives of

Emergency Medicine, 4, 77±82.

Fergusson, E., Li, J., & Taylor, B. (1998). Grandmothers' role

in preventing unnecessary accident and emergency attend-

ances: Cohort study. British Medical Journal, 317, 1685.

Frew, S. (1991). In Patient screening requirements in patient

transfers: How to comply with the law (pp. 7±10). Dallas,

TX: American College of Emergency Physicians.

Green, J., & Dale, J. (1992). Primary care in accident and

emergency and general practice: a comparison. Social

Science & Medicine, 35, 987±995.

Grumbach, K., Keane, D., & Bindman, A. (1993). Primary

care and public emergency department overcrowding.

American Journal of Public Health, 83(3), 372±378.

Haddy, R., Schmaler, M., & Epting, R. (1987). Nonurgent

Emergency room Use in Patients with and without

Primary Care Physicians. Journal of Family Practice, 24(4),

389±392.

Hansagi, H. (1990). Referral of non-urgent cases from an

emergency department: patient compliance, satisfaction

and attitudes. Scandinanvian Journal of Social Medicine,

18, 249±255.

Hay, J. (1992). In Health care in Hong Kong (pp. 10±11).

Chinese University Press Chapter 1.

Hull, S. A., Jones Rees, I., & Moser, K. (1997). Factors in¯u-

encing the attendance rate at accident and emergency

departments in East London: the contributions of practice

organization, population characteristics and distance.

Journal of Health Services Research and Policy, 2(1), 6±13.

Klijakovic, M., Allan, B. C., & Reinker, J. (1981). Why skip

the General Practice and go to the Accident and

Emergency Department? NZ Medical Journal, 93, 49±52.

Lee, A. (1997). Family medicine in Hong Kong: Present and

future. Letter from Hong Kong. Canadian Family

Physician, 43, 1049±1052.

Lee, A., Lau, F. L., Hazlett, C. B., Kam, C. W., Wong, P.,

Wong, T. W., & Chow, S. (1999). Measuring the inap-

propriate utilization of Accident and Emergency services.

International Journal of Health Care Quality Assurance,

12(7), 287±292.

Lee, A., Lau, F. L., Hazlett, C. B., Kam, C. W., Wong, P.,

Wong, T. W., & Chow, S. (1998). Why patients with non

urgent conditions utilise Accident and Emergency

Services? Can patients or nurses classify their severity of

illness? In Building a New Health Care System: A joint

scienti®c conference hosted by The Institute for Health

Policy and Systems Research and The Hong Kong Chapter

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±10851084

of the Society of Alumni of The Johns Hopkins University

School of Hygiene and Public Health. 15±16 June 1998,

Hong Kong.

Leicester, M. C., Johnston, J., Daniel, L., Ng, N. Y., Hedley,

A. J., Fung, H., Leung, M. P., Chung, C. H., 1991. Hong

Kong Accident & Emergency Departments: Demand,

workloads & outcomes. Department of Community

Medicine, University of Hong Kong & Hospital Services

Department, Hong Kong Government.

Morris, F., Head, S., & Holkar, V. (1989). The nurse prac-

titioner: Help in clarifying clinical and educational activi-

ties in Accident and Emergency Departments. Health

Trends, 21, 124±126.

Murphy, A. W., Plunkett, P. K., Bury, G., Lynam, Smith,

M., & Gibney, D. (1987). E�ect of the introduction of a

®nancial incentive for fee-paying A&E attenders to consult

their general practitioner before attending the A&E

department. Family Practice, 14(6), 407±410.

Myers, P. (1982). Management of minor medical problems

and trauma: General Practice or hospital? Journal of the

Royal Society of Medicine, 75, 879±893.

National Centre for Health Statistics (McGraig LF), 1994.

National Hospital Ambulatory Medical Care Survey:

1992 emergency department summary. Advance data

from vital and health statistics. No 245. Hyattsville, MD:

Public health Service, (DHHS publication no. (PHS) 94-

1250).

O'Grady, K. F., Manning, W. G., Newhouse, J. P., & Brook,

R. H. (1985). The impact of cost sharing on emergency

department use. New England Journal of Medicine, 313,

484±490.

Padgett, D., & Brodsky, B. (1992). Psychosocial factors in¯u-

encing non-urgent use of the emergency room: a review of

the literature and recommendations for research and

improved service delivery. Social Science & Medicine,

35(9), 1189±1197.

Selby, J. V., Fireman, B. H., & Swain, B. E. (1996). E�ect of

a copayment on use of the emergency department in a

health maintenance organization. New England Journal of

Medicine, 334, 635±641.

Shah, N., Shan, M., & Behbehani, J. (1996). Predictors of

Non-urgent Utilization of Hospital Emergency services in

Kuwait. Social Science & Medicine, 42(9), 1313±1323.

Smith, J., 1995. Focus: London: Threats and opportunities in

accident and emergency (Editorial), 311, 7018, p. 1456.

Vayda, E., Gent, M., 1973. An emergency department triage

model based on presenting complaints. Emergency Triage

Model.

A. Lee et al. / Social Science & Medicine 51 (2000) 1075±1085 1085