Factors influencing the attendance rate at accident

Journal of Health Services Research http://hsr.sagepub.com/

Factors Influencing the Attendance Rate at Accident and Emergency Departments in East London: The Contributions of Practice Organization, Population Characteristics and Distance Sally A. Hull, Ian Rees Jones and Kath Moser J Health Serv Res Policy 1997 2: 6 DOI: 10.1177/135581969700200104 The online version of this article can be found at: http://hsr.sagepub.com/content/2/1/6

Published by: http://www.sagepublications.com

Additional services and information for Journal of Health Services Research can be found at: Email Alerts: http://hsr.sagepub.com/cgi/alerts Subscriptions: http://hsr.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Jan 1, 1997 What is This?

Downloaded from hsr.sagepub.com at The Royal Society of Medicine Library on October 8, 2014

Original research

Factors influencing the attendance rate at accident and emergency departments in East London: the contributions of practice organization, population characteristics and distance Sally A. Hull. Ian Rees Jones*. Kath Moser Department of General Practice, St Bartholomew's and the London Medical College; *Department of Geography, Queen Mary and Westfield College, London, UK - - - - - -------- - - - - - - - - - - - - - -

Objectives: To examine the contribution of general practice organisation, population characteristics and distance to practice attendance rates at four local accident and emergency departments. Design: Practice-based study examining variations in accident and emergency department attendance rates in 105 practices, using routine data from the Family Health Services Authority (FHSA), the District Health Authority and the 1991 Census. Setting: East London and the City Health Authority, covering practices based in the inner city boroughs of Hackney, Tower Hamlets and Newham, and the City of London. Main outcomemeasure: Practice-based, age-standardized, adult attendance rates at accident and emergency departments in the year to 31 March 1994. Results: Annual age-standardized practice accident and emergency department attendance rates ranged from 10.3 to 29.4 per 100 population. The mean practice attendance rate was 17.6 per 100 (95% CI 16.8-18.4). No significant relationship was found between attendance rates and practice characteristics (number and sex of general practitioner (GP) principals, presence of practice manager or nurse, computerization and training status). There were strong positive relationships between attendance rates and households not owner-occupied (R = 0.55, P < 0.001) and pensioners living alone (R =0.55, P < 0.001). There were negative correlations with Asian ethnicity (R =-0.31, P =0.002) and residents lacking amenities (R =-0.26, P =0.007). The distance to the nearest accident and emergency department also correlated negatively with attendance (R = -0.27, P = 0.006). A backwards multiple regression model showed that 48% of the variation in attendance rates could be accounted for by six factors: percentage of households not owner occupied, percentage living in households without a car, percentage living in households lacking amenities, percentage of pensioners living alone, percentage of Asian ethnicity, and percentage living in households with a head bom in the New Commonwealth and Pakistan. Optimal subsets regression identified a number of altemative models with similar explanatory value. Conclusions: Social deprivation is strongly linked with attendance rates at accident and emergency departments in East London. In contrast, the organizational characteristics of general practices appear to have no bearing on the rates. Both purchasers and providers need to take account of these f"mdingswhen planning accident and emergency provision. Journal of Health Services Research and PolieyVoI. 2 No. I, 1997: 6--13

Introduction Accident and emergency departments (AEDs) are an important interlace between primary and secondary care, with considerable overlap in the problems pre--------------------------

Sally A. Hull, Senior Lecturer, Department of General Practice, St Bartholomew's and the London Medical College, London £1, UK Ian Rees Jones, Lecturer, Department of Geography, Queen Mary and Westfield College, London £1, UK Kath Moser, Research Officer, Department of General Practice, St Bartholomew's and the London Medical College, London £1, UK Correspondence to I. R. J.

6

© Pearson Professional Ltd 1997

sented in both settings.' Over the past 30 years there has been a steady rise in AED attendance rates, from 105 per 1000 population in 1961 to 241 per 1000 in 1991. 2,3 This rise in use is greatest in urban areas." Little is known about the main determinants of this increasing use. Purchasers of health care, whether GP fundholders or health authorities, have an interest in attendance rates as they predict the use of other hospital services. Both in inner London and outside London one in five AED attenders will go on to be admitted," and 10% will be referred on to outpatients." Day-time patterns of attendance at AEDs mirror GP

J Health Serv Res Policy Volume 2 Number 1 January 1997 Downloaded from hsr.sagepub.com at The Royal Society of Medicine Library on October 8, 2014

Original research

Factors influencing attendance rate at AEDs in East london

availability in surgery, with peaks in the morning and early evening." The majority of attenders are self-referrals; general practice initiated referral rates are similar both within inner London and outside." Estimates ofattenders whose problems could be managed within primary care range from 35% to 70%; comparison between studies is hampered by the variety of ways that 'appropriate for primary care' has been defined by different authors. Responses to this rise in demand fall into two categories, which are illustrated within the Tomlinson report. B Whilst encouraging schemes that develop the role of AEDs as primary care providers, the report - alongside other descriptive studies - also suggests that increasing the availability of well-organized primary care, particularly in inner London, could reduce the burden on AEDs. B,9 This assumes a relationship between organizational features of general practice and AED attendance rates. However, there is a lack of clear evidence on the factors affecting use of AEDs. No previous practice-based study has demonstrated an association between patient or practice characteristics and attendance rates. lO,11 Some studies, set in rural areas, have shown a significant association between attendance at AEDs and distance from the patient's home, but this association has not been studied within inner city areas. 1O•12 This study, using routine data from the health authority and 1991 Census, investigates the relationship between attendance at AEDs in East London and characteristics of practice organisation, population and distance to the AEDs. It therefore offers new insights into the relative importance of these different factors as predictors of attendance, and hence whether planned changes within practices might result in altered AED attendance rates.

Method Setting East London and the City Health Authority (ELCHA) is a deprived, multi-ethnic, inner city area in East London, co-terminous with the boroughs of Hackney, Newham, Tower Hamlets and the City of London. The study was conducted using routinely collected data from the City and East London Family Health Services Authority (FHSA) , the health authority (ELCHA), and the 1991 Census. Attendances, between April 1993 and March 1994, at four AEDs (The Royal London, St Bartholomew's, Homerton and Newham) were obtained for ELCHA residents registered at all practices in the district. Children were excluded from the analysis because of their use of Queen Elizabeth Hospital (a special health authority hospital for which no comparable data were available). There were approximately 86 000 attendances that could be linked to east London practices. As this is a practice-based study, our analysis excluded attendances by patients who had no recorded GP; these include the homeless and refugees who have complex health needs and make up London's invisible population.P

Practice and population characteristics The City and East London general practice database was used to provide information on practice characteristics. 14 This included the number and sex of principals, the presence of a nurse, practice manager, computer and general practice trainer (data collected between June and October 1993). These variables, collected from routine data sources, were included as the best available indicators of practice organization, innovation, skill mix and facilities. Practice population characteristics were calculated using the postcode distribution of the practice lists (for summer 1993) applied to the 1991 Census data by ward. The linear distance between each practice and the four AEDs included in the analysis was calculated from grid references using Arc Info software, 15 and the distance in kilometers to the nearest AED was used in our analysis. The full list of variables used in the analysis is given in Table 1.

Practices retained for analysis We restricted our analysis to practices where we believed the attendance data to be comprehensive and reliable. Hence practices where 10% or more of adult admissions via AEDs were to hospitals outside the ELCHA area were excluded from the analysis. This eliminated border practices whose patients frequented AEDs outside the district and for which attendance data were not available. Practices where 20% or more of the list lived outside the ELCHA area were also excluded in order to minimize bias in the calculation of standardized attendance rates (see below). Shared-site practices were excluded because of uncertainty as to whether attenders were allocated to the correct practice. Of the remaining 107 practices, information on population characteristics was unavailable for one, which was excluded. One further practice was excluded: at 62.4 per 100, its AED attendance rate was over 10 standard deviations above the mean attendance rate used in our final analysis (17.6 per 100). Of the total 163 practices in ELCHA, 105 remained for analysis (approximately 63000 attendances). No significant differences in practice characteristics were found between the included and excluded practices (using X2 tests).

Calculating a standardized attendance rate for each practice An age-standardized attendance rate per 100 population was calculated for each practice. Direct standardization was used taking as the standard population the 1991 mid-year estimate (residents only) of East London and the City.16 The practice age distributions for ELCHA residents were obtained by applying the proportion of the practice list (all ages) resident in ELCHA to the practice age distributions of all patients, wherever resident. Using these estimated age distributions, AED attendance rates per 100 population were calculated for age groups 15-44, 45-64, 65-74 and 75 and over.


7

Original research Table 1

Factors influencing attendance rate at AEDs in East London

Practice and population variables used in the analysis

Practice characteristics Number of general practitioner (GP) principals Presence of a female GP principal Presence of a nurse (practice employed or health promotion) Presence of a practice manager Presence of a GP trainer Presence of a computer Distance Distance (km) from the practice to the nearest accident and emergency department (AED) Population variables Male unemployment Female unemployment Permanent sickness SEG 11 SC 1V or V Male sickness Female sickness Asian ethnicity Black ethnicity Other ethnicity White ethnicity Born in NCP Migration Lone parent household Pensioners alone Not owner-occupied No car Lacking amenities Overcrowding

% of economically active males (16+) unemployed % of economically active females (16+) unemployed % permanently sick (16+) % living in households whose head is in socio-economic group (SEG) 11 (unskilled manual) % living in households with economically active head in social class (SC) IV or V (semiskilled or unskilled) Standardised limiting long-standing illness: males 0-74 Standardised limiting long-standing illness: females 0-74 % whose ethnicity is Asian % whose ethnicity is Black % whose ethnicity is Chinese or 'other' % whose ethnicity is White % living in households headed by person born in New Commonwealth or Pakistan (NCP) % who have changed address in the past year % living in one parent households

% of pensioners living alone % of households not owner-occupied % living in households without a car % living in households lacking or sharing amenities

% living in households with more than one person per room

Statistical analysis

Table 2 Age-standardized adult AED attendance rates per 100 population in 105 practices in East London

The unit of analysis throughout was the general practice. Univariate relationships between practice characteristics and attendance rates were examined by calculating mean standardized attendance rates (and 95% confidence limits (CIs)) for appropriate subgroups of practices. For population and distance variables we calculated the Pearson's product moment correlation coefficients between these variables and attendance rates, and P values obtained from a two-tailed test of significance. Multiple regression analyses were used to examine the relationship between population and practice characteristics and attendance rates. Backwards stepwise multiple regression and optimum subsets multiple regression were both undertaken using population and practice characteristics as explanatory variables and the standardised attendance rate as the dependent variable. All variables listed in Table 1 were entered in the backwards regression satisfying an in-eriterion of P = 0.045; they were removed using an out-criterion of P = 0.05. The analysis was undertaken using SPSS17 and BMDp 18 software.

Results Table 2 shows the considerable vananon in practice AED attendance rates; the mean value was 17.6 (95% CI 16.8-18.4) per 100 adult population with a range of 10.3-29.4.

Univariate analysis Practice characteristics and attendance rates No significant relationship was found between AED

8

Attendance rate per 100 population

Number of practices

10-11.99 12-13.99 14-15.99 16-17.99 18-19.99 20-21.99 22-23.99 24-25.99 26-27.99 28-29.99

5 18 19 17 14 20 5 2 3 2

105

attendance rates and any of the practice characteristics used in the analysis (number of principals, presence of a female principal, nurse, practice manager, computer, GP trainer) (Table 3).

Population characteristics, distance to AED and attendance rates Table 4 shows the variation between practices in their population characteristics, and the correlation between these variables and AED attendance rates. The strongest positive relationships were with households not owneroccupied (R = 0.55, P< 0.001) and pensioners living alone (R = 0.55, P< 0.001). There were negative correlations with Asian ethnicity (R= -0.31, P= 0.002) and residents lacking amenities (R = -0.26, P = 0.007). The distance from the practice to the nearest AED showed a negative correlation with attendance rates (R = -0.27, P= 0.006).


Original research

Factors influencing attendance rate at AEDs in East London Table 3

Mean annual practice standardized AED attendance rates by practice characteristics No. of practices

Practice characteristics

Mean AED attendance rate per 1DO population

95% CI

- - - - - - - --_.

-----------

Number of principals

1 2or3 4 or more

39 46 20

17.5 17.6 18.1

16.0-19.1 16.4-18.7 16.3-19.9

No. of female principals

~

1 None

44 61

17.3 17.8

16.2-18.5 16.7-18.9

No. of practice managers

~

1 None

53 52

17.6 17.7

16.5-18.7 16.5-18.9

No. of nurses

~

1 None

66 39

17.3 18.2

16.3-18.2 16.8-19.7

Training practice

Yes No

14 91

18.4 17.5

16.3-20.4 16.6-18.4

Computer

Yes No

64 41

17.5 17.8

16.5-18.6 16.5-19.1

105

17.6

All practices

~~--_._-_._--------------._----_._------ - - - - - - -

16.8-18.4 -

----~---

Table 4 Mean values of population and distance variables and correlations between these and the AED attendance rates Mean percentage (range) of variables"

Variable" -~-----~--~-_.

----

Population variables Pensioners alone Not owner occupied No car Male sickness Male unemployment Female sickness Permanent sickness Other ethnicity Lone parent household Asian ethnicity Lacking amenities Female unemployment White ethnicity Born in NCP Black ethnicity SEG II Migration SC IVorV Overcrowding

Correlation coefficient'

----------

39.2 (32.5--44.6) 65.5 (36.1-84.1) 51.4 (36.6-64.7) 131.6 (112.5-153.0) 24.6 (17.3-31.2) 137.9 (119.9-166.7) 5.1 (3.9-6.8) 2.8 (1.7--4.0) 8.2 (4.1-14.7) 19.6 (4.6-50.0) 2.1 (0.3-3.8) 16.1 (11.7-20.9) 63.3 (36.4-81.1) 30.0 (14.2-57.1) 14.3 (4.8-27.8) 4.4 (2.3-8.4) 10.6 (7.9-13.5) 26.6 (20.0-33.9) 18.9 (10.4--44.4)

Distance variable Distanceto nearest AED (km)

1.8 (0.1--4.7)

0.55*** 0.55*** 0.48*** 0.48*** 0.47*** 0.35*** 0.35*** 0.35*** 0.32** -0.31 ** -0.26** 0.22* 0.21 * -0.21* 0.20* 0.14 0.11 0.09 0.09 -0.27**

Details of variables are given in Table 1. Exceptfor male and female sickness which are mean standardized ratios (England and Wales = 100). e Two-tailed test of significance: * P < 0.05, ** P < 0.01, ***P < 0.001. a

b

Multiple regression analysis Table 5 gives a selection of models produced by the optimal subsets method showing those with the two highest If values for models with four, five and six variables. The model for the backwards stepwise regression is also summarized; it is the same as one of the six variable optimal subsets models. Table 5 illustrates that there is a range of alternative models with similar predictive value, with an If of between 0.44 and 0.48. The variables of lacking amenities, no car and not owneroccupied feature in all the models shown. Table 6 shows the results of the backwards elimination model in more detail. Once the six variables shown have been taken into account, none of the other variables entered into the model are important in explaining the

attendance rates. The multiple correlation coefficient is 0.70, with these six variables accounting for 48% of the variation in AED attendance (F(6,98) = 15.30, P< 0.0001). The residuals are normally distributed. It can be seen that for three variables (no car, lacking amenities, and Asian ethnicity) the direction of the relationship with AED attendance changes as a result of being in the model. Table 7 shows the high correlation coefficients that exist between many of the variables in the backwards model.

Discussion This study suggests that the demand for care at AEDs is associated with the social characteristics of the population, while organizational features within general prac-

---------------


9

Original research Table 5


A selection of the best multiple regression models obtained via the optimal subsets and backwards regression methods. Optimum subsets"

Variables in each model

Four variables Adjusted

R' R'

0.440 0.417

Five variables

0.423 0.400

0.464 0.437

-.J -.J -.J -.J

-.J -.J -.J

• • •

-.J

-.J -.J -.J -.J -.J

• •

• •

._.

Lacking amenities No car Not owner-occupied Asian ethnicity Female unemployment Pensioners alone Born in New Commonwealth

0.460 0.433 --

•

Backwards stepwise Six variables

0.484 0.452

--------

0.481 0.449

0.484 0.452

----------------------

-.J -.J -.J -.J

-.J -.J -.J -.J

•

•

-.J

-.J -.J

•

-.J -.J -.J -.J

-.J -.J -.J -.J -.J -.J

• -.J -.J

•

--- -

-----

-.J, in the model; ., not in the model. a

Two highest R' in subsets with four, five and six variables.

Table 6 Backwards regression of population and practice variables on standardized attendance rate at AEDs (variables in the equation only) Variable

8

Pensioners alone Not owner-occupied No car Lacking amenities Asian ethnicity Born in New Commonwealth (Constant)

0.6 0.95 -1.24 5.40 0.37 --D.24 -16.04

95% CI 0.05 0.61 -1.75 3.07 0.14 --D.46 -31.20

p values (for T)

1.18 1.28 -0.73 7.74 0.60 -0.01 -0.88

0.0331 0.0000 0.0000 0.0000 0.0016 0.0382 0.0384

R'

Adjusted R'

= 0.48 = 0.45

R = 0.70

B = unstandardised regression coefficients

Table 7

Correlations between the variables in the backwards model Pensioners alone

Pensioners alone Not owner-occupied No car Lacking amenities Asian ethnicity Born in New Commonwealth

0.85 0.83 --D.46 --D. 53 --D.41

Not owner-occupied

0.91 -0.75 -0.47 -0.44

tices do not appear to influence the rates. There is a negative association between distance from the nearest AED and attendance rates. Both univariate and multivariate analyses indicate that there are likely to be higher attendance rates in populations where owner occupation is low and where there is a high proportion of pensioners living alone, both of which are generally taken to be indicators of social deprivation. However, the relationships between some of the other population characteristics identified as important and attendance rates are not straightforward, as indicated by the high correlations between some of these variables (Table 7) and the change in direction of the regression coefficients of three of the variables (no car, lacking amenities and Asian ethnicity) between the univariate and multivariate analyses. Clearly complex interactions between several of the variables exist, and this is an area which may merit further investigation with more sophisticated information than we were able to obtain using routine data. Nevertheless, it appears from this study that some aspects of social deprivation are influential in determining the demand for accident and emergency (A&E) care.

10

No car

--D. 52 -0.19 -0.13

Lacking amenities

Asian ethnicity

0.38 0.57

0.90

Born in New Commonwealth

Practice characteristics and distance to AEDs We found no evidence of a relationship between the standardized practice AED rates and any of the routinely collected practice characteristics available on the database. Hence there was no evidence to suggest that practice size, staffing levels and training status had any influence on use of AED by registered patients. This finding is important because it contrasts with evidence from studies using similar sources of routinely collected data suggesting a link between staffing levels, training status and practice-based performance such as appropriate asthma prescribing and achievement of targets for cervical cytology.19,20 It could be argued that the variables entered on the database from routine sources are not subtle enough to capture the potentially important practice differences in the delivery of care which influence AED attendance, such as the workings of practice appointment systems or ease of contact by telephone. Availability of out-of-hours care may also be important; however, studies have shown that only 3-6% of patients attempt to contact their GP before attending anAED. 2



Original research

The availability of GPs to their patients, mediated by the practice appointment system, is an important potential influence on attendance at AEDs. This was examined by Campbell in a prospective study of 19 practices in West Lothian which also sought patients' views on availability.'! The study measured provision of appointments, appointment availability at the start of the day, and numbers of 'extra' patients seen (all adjusted for list size). Whilst these measures of availability varied widely between practices, and were associated with dissatisfaction with the perceived availability for non-urgent consultations, no association was found with AED selfreferral rates. It is possible that practice organizational features other than those we were able to measure influenced AED rates. At present, GPs are unaware of the magnitude of AED use among their patients or of factors that might influence AED self-referral behaviour, and, at present, there is little incentive to develop mechanisms that might affect these rates. Whilst it is tempting to compare this situation with the provision of cervical cytology screening before and after the introduction of targets, there is as yet no evidence that practice-based interventions affect self-referral rates to AEDs. We found an association between AED attendance rates and distance, even though the distance between practice and the nearest AED was calculated 'as the crow flies' and took no account of ease and means of transport, or the geographical distribution of practice lists. Using routine data, it was not possible to examine whether those people living furthest from their practice surgery are more likely to go to an AED, or indeed whether the choice between competing local AEDs is made on the basis of distance alone or on other factors such as waiting times and previous experience. Our study confirms the association between attendance rates and distance to AEDs, even within a densely populated inner urban area where several competing AEDs may be within easy reach of the practice and its population.

interest in developing such mechanisms, particularly in areas of social deprivation where the burden of attendance is greatest. We have used data that were routinely collected for other purposes. The advantage of doing this is that the time spent on data collection is minimal, and it puts to good use the vast amounts of data generated by general practice and the health authority for administrative and financial purposes. The major limitations are that the data may be of uncertain accuracy and are not tailormade for our requirements. For example, a prospective study would have enabled collection of a wider range of practice-based variables. The reliability and completeness of the AED data are unclear. In an inner city area with high practice turnover and large homeless populations, a patient's GP will sometimes be incorrectly recorded or not recorded at all. GP lists in east London are often inaccurate and 'list inflation' may run at 20-30%.23 Inflated denominators will give rise to deflated attendance rates. List inflation may be a confounding factor in that practices with high list inflation may have high levels of social deprivation. This will only serve to underestimate the differential in AED attendances. Our dataset related to AED episodes and there were no data on the number of recurrent AED visits. There may be potential patient clustering that is not taken account of in our model. Census-derived variables for general practices have previously been used to examine cervical smear uptake and night visit rates. 20,24 The accuracy of such variables has been called into question, and there is no doubt that they should be treated with caution." It has been suggested that variables derived from enumeration districts are more accurate than ward-based statistics." The calculation of these practice population variables rests on certain assumptions: firstly, that ward populations are homogeneous which is patently not the case in east London where very different types of population groups live side by side; secondly, that patients registered at a practice are representative of the ward in which they live. However, despite the problems associated with using imputed practice population variables, this is the basis on which capitation deprivation payments are made to practices and therefore has great practical importance. If anything, the variability between practices will be attenuated by the use of variables derived in this way. Emerging from these results is a further set of questions which should inform future research. These include the impact of closing a unit and the effect of altered out-of-hours provision in general practice on AED attendance. Of particular interest for east London, and areas with similar populations, are the varying relationships of attendance rates with different ethnic groups. The ecological fallacy is ever present in such data, but this finding might relate to important differences in access to care, to differentials in morbidity, or to differences between cultures in the threshold of seeking emergency care.

Strengths and limitations of the analysis By including in our analysis attendances at the four AEDs in east London over a 12-month period, we have been able to capture a district-wide view of an activity which is of great importance for both purchasers and providers of health care. Understanding some of the predictors of high AED rates is of particular interest to practices engaging in total purchasing pilot projects where practices are given a budget to cover all their patients' use of services, including A&E. High AED attendance rates are associated with high rates of admission and referral onwards to outpatient departments. This may stimulate the development of mechanisms for regaining control over this 'gateway' into secondary services. For example, recent studies have shown that GP management of non-emergency patients in AEDs results in reduced rates of investigation and referral without any apparent detriment to outcomes. 21,22 Highvolume health authority purchasers should have an


11

Original research


Getting AED provision right for the local population The reasons for AED attendance are complex and are related to patients' perceptions of the purpose of AEDs as well as to differentials in morbidity and supply factors. 27,28 A qualitative study by Holohan/" showed how accidents are seen by individuals as isolated episodes requiring treatment rather than diagnosis. In her study, satisfaction with the technical competence of AEDs did not indicate dissatisfaction with GPs as the entry system for care and referral for illness, If, as Holohan argues, patients suffering accidents view their condition as serious because of the interruption of social roles, then these public views may be in conflict with attempts to define appropriateness of location and time of care in medical terms. Recent reviews of the literature suggest that demand for primary care in AED settings will persist in urban areas and that attempts to deflect this demand into community settings will have little success. 3D In the light of our study, attempts to influence AED attendance rates by improving general practice organization may be misguided. Whilst health care policy should be directed at developing and evaluating specific mechanisms to address AED use, attention should not be diverted from the underlying importance of social deprivation.

Acknowledgments We are grateful to the City and East London FHSA for providing the practice data. We thank Jacqui Bobby and Malcolm Palmer at ELCHA information department for downloading and checking the attendance data, Ian Gregory at Queen Mary and Westfield College Department of Geography for manipulating the grid reference data, Dr Susan Dolan, Department of General Practice, St Mary's Hospital Medical School, and Mike Chambers, Department of Epidemiology and MedicalStatistics, St Bartholomew'sand the Royal London School of Medicine and Dentistry, for providing us with the population variables for the practice populations, and Sandra Eldridge, also of the Department of Epidemiology and Medical Statistics, for statistical advice.We are grateful to the Cityand East London database project. The database project was supported by the City and East London FHSA and the former North East Thames Regional Health Authority. The socio-demographic aspect of the work was funded by the North East Thames Regional Health Authority locally organized research scheme. Conflict of interest: None.

References 1. DaleJ, GreenJ, Reid F, Gluckman E. Primary care in the accident and emergency department: British Medical Journal 1995;311:423-426 2. Hallam L. Primary medical care outside normal working hours: review of published work. British Medical Journal 1994;308:249-253 3. National Audit Office. NHS Accident and emergency departments in England. London: HMSO, 1992 4. Milner P C, NichollJ P, Williams B T. Variations in demand for accident and emergency departments in England from 1974 to 1985. Journal of Epidemiology and Community Health 1988; 42: 274-278 5. Jankowski R F, Mandalia S. Comparison of attendance and emergency admission patterns at accident and emergency departments in and out of London, British Medical Journal 1993;306: 1241-1243 6. Reilly P. Primary care and A&E departments in an urban

12

area. Journal of the Royal College of General Practitioners 1981;31:223-230 7. Farmer D T, ChambersJ D. The relationship between accident and emergency departments and the availability of general practitioner services - a study in six London hospitals. London: King's Fund, 1982 8. Tomlinson B. Report of the inquiry into London's health services, medical education and research. London: HMSO,1992. 9. Davison A G, Hildrey A C C, Floyer M A, Use and misuse of an accident and emergency department in the east End of London. Journal of the Royal Society of Medicine 1983; 76(1): 37-40 10. McKee C M, Gleadhill D N S, WatsonJ D. Accident and emergency attendance rates: variation among patients from different general practices. BritishJoumal of General Practice 1990; 40: 150-153 11. CampbellJ L. General practitioner appointment systems, patient satisfaction, and use of accident and emergency services - a study in one geographical area. Family Practice 1994; 11: 438-445 12. Ingram D R, Clarke D R, Murdie RA. Distance and the decision to visit an emergency department. Social Science and Medicine 1978; 12: 55-62 13. Jacobson B. Public health in inner London. In: SmithJ (ed.) London after Tomlinson, reorganising big city medicine. London: British MedicalJoumal Publications, 1993 14. NaishJ, Sturdy P, Toon P. Appropriate prescribing in asthma and its related cost in East London. British Medical Journal 1995; 310: 97-100 15. Arc Info Version 6. (1992). Environmental Systems Research Institute Inc., Redlands, CA, USA 16. East London and the City Health Authority. Census manual. East London and the City Health Authority, 1993 17. SPSS for Windows, Version 6.0. Chicago: SPSS Inc., 1993 18. BMDP Statistical Software, software release. California: BMDP Statistical Software Inc., 1990 19. Sturdy P, NaishJ, Pereira F, Griffiths C, Dolan S, Toon P et al. Characteristics of general practices that prescribe appropriately for asthma. British MedicalJoumal1995; 311: 1547-1548 20. Majeed F A, Cook D G, Anderson H R, Hilton S, Bunn S, Stones C. Using patient and general practice characteristics to explain variations in cervical smear uptake rates, British Medical Journal 1994; 308: 1272-1276 21. Dale ], Lang H, RobertsJ, GreenJ, Glucksman E. Cost effectiveness of treating primary care patients in accident and emergency: a comparison between general practitioners, senior house officers and registrars. British Medical Journal 1996; 312: 1340-1344 22. Murphy A W, Bury G, Plunkett P K, Gibney D, Smith M, Mullan E et al. Randomised controlled trial of general practitioner versus usual medical care in an urban accident and emergency department: process, outcome and comparative cost. British Medical Journal 1996; 312: 1135-1142 23. RobsonJ, Falshaw M. Audit of preventive activities in 16 inner London practices using a validated measure of patient population, the 'active patient' denominator. British Journal of General Practice 1995; 45; 463-466 24. Majeed F A, Cook D G, Hilton S, PolonieckiJ, Hagen A. Annual night visiting rates in 129 general practices in one family health services authority: association with patient and general practice characteristics. BritishJoumal of General Practice 1995; 45; 531-535 25. Majeed F A, Cook D G, PoloneickiJ, GriffithsJ, Stones C. Sociodemographic variables for general practices: use of census data. British MedicalJoumal1995; 310: 1373-1374 26. Scrivener G, Lloyd D C E F. Allocating census data to general practice populations: implications for study of prescribing variation at practice level, British Medical Journal 1995;311: 163-165



27. Singh S. Selfreferral to accident and emergency departments: patient's perceptions. British Medical Journal 1988;297: 1179-1180 28. Eachus], Williams M, Chan P, Davey Smith G, Grainge M, Donovan] et al. Deprivation and cause specific morbidity: evidence from the Somerset and Avon survey of health. British Medical]ournaI1996; 312: 287-292

Original research

29. Holohan A M. Accident and emergency departments: illness and accident behavior. The sociology of the NHS: Sociological Review Monograph, 1976; 22: 111-119 30. Green], DaleJ. Primary care in accident and emergency and general practice: a comparison. Social Science and Medicine 1992; 35: 987-995


13