Methods

sampling

The study was undertaken between October and December 1997 in the A&E department of a major teaching hospital in Sheffield, England. The A&E department is an adult facility, receiving approximately 80 000 new attenders each year. It is typical of other medium sized A&E departments located in cities in England. On arrival, all new attenders are prioritised for treatment to colour coded triage streams. This A&E operates a five colour system (black, red, blue, green, and yellow). The two indicating the lowest priority for treatment are “green”, classified broadly as “new illnesses or injuries which are non-urgent”, and “yellow”, “injuries or long-standing problems where advice could have been sought elsewhere”. In the study period 8057 adult attenders were assigned to either the “green” or the “yellow” treatment streams. Together these two low priority for treatment groups accounted for 61.5% of the all new A&E attenders, and formed the sample frame for the survey.

A sample size of 150 from each of the “yellow” and the “green” treatment streams was estimated to be sufficient to enable the percentage of patients in each stream who were treatable by services other than A&E to be calculated within +/- 5% assuming that the true proportion is 10%. Using a simple Latin square design (7 × 7 × 7), the A&E nurse recruited the first three patients assigned to each of the “yellow” and the “green” treatment streams in each of seven pre-selected “sessions”. The start times of the sessions rotated daily through 24 hours for seven consecutive weeks. The design yielded a target sample of 147 patients from each of the “yellow” and “green” streams, totalling 294. The inclusion criteria were that participants should not be incapable through alcohol or drugs to complete a questionnaire, or have been referred to A&E formally by letter from their general practitioner.

data collection

Data for the study were collected by a questionnaire completed by the patient shortly after arriving in A&E. The A&E notes were then reviewed by a member of the research team. Ethical approval for the study was obtained from the local ethics committee, and signed consent to access the notes was given by each patient.

Patient perceptions of their health problem, how, where and when it occurred, and sociodemographic indicators were collected on the questionnaire. The most frequently quoted reasons for attending A&E published in the literature^12–14 were configured into 30 statements (table 1). Participants were asked to indicate from the reasons listed all those that had been instrumental in their decision to attend A&E and also to add any “other” reasons. Information about referral, diagnoses, processes of care and disposal was collected from the A&E notes. The diagnoses in the notes were coded to the International Classification of Diseases 9th revision (ICD9).¹⁵ All the health problems within the range ICD9 800–999 (injury and poisoning) [ibid:735] were recoded as “trauma” and all else was recoded as “non-trauma”.

analysis

The data were coded and analysed using SPSS for Windows. Population estimates of the total attendance at A&E with less urgent conditions were calculated after weighting the numbers in the sample recruited from the “green” and “yellow” streams to reflect their proportions in the sample frame (92% “green” and 8% “yellow”) and 95% confidence intervals were calculated for key estimates. To enable the effect of the weighting to be assessed, unless otherwise stated, all the numbers reproduced in this paper are raw data and all the population estimates are based on weighted numbers.

suitability for treatment by other providers

The patients were classified into subgroups based on their suitability for treatment in other settings (box 1). The alternative models of healthcare considered were general practice, a nurse led minor injury unit (MIU), a walk in centre, with the capacity to treat both minor injuries and minor medical conditions, and the national telephone advice line NHS Direct that can advise on the management of minor health problems not needing medical care.

The criteria we used to identify patients suitable for treatment in general practice were developed together with GPs in a UK study.¹⁶ The criteria to identify suitability for treatment in the MIU were developed by applying the nurse practitioner protocols in use at the MIU located at another hospital site in Sheffield to both the description of the health problem given by the patient on the questionnaire, and the processes of care received as recorded in the notes. Walk in centres will have various specifications¹⁰ but the model we assumed to distinguish it from the hospital-based MIU was on a site remote from an A&E department (for example in a railway station or high street setting) staffed by nurses working within protocols similar to those used in the MIU that permit some treatments, advice, and prescribing to be carried out, but that unlike the MIU do not have direct access to radiography facilities. The group identified as appropriate for NHS Direct were those whose processes of care defined them as suitable to receive advice to “self care” (box 1). The group identified as suitable for treatment in a traditional A&E setting was the “default” category so defined if none of the criteria for suitability for treatment or advice in the other models of care were met.

reasons for the patient's decision to attend a&e

We classified the reasons for attending A&E given by the patients on the questionnaire into “strong”, “medium” and “weak” (table 1). “Strong” reasons for attending A&E were those we considered to have low potential for changing consulting behaviour. Reasons scored as “medium” were those which offered moderate potential for changing the decision to attend A&E. The reasons graded as “weak” were those we considered to offer the highest potential for changing the decision to attend A&E.

The reasons, “seeking a radiograph”, (seeking particular A&E services) or “I consider the condition as emergency” (perceptions of seriousness), were classified as “strong” reasons. Reasons reflecting positive experiences of A&E in the past were also graded as “strong” in that they reflected previous patterns of consulting behaviour that might be difficult to influence. The reason being “...advised to attend by someone else” was classified as “medium” if the source of the advice was “friends or family”, and “strong” if the advice came from “others” in a formal relationship with the patient, such as a works nurse. Reasons reflecting “convenience of A&E services”, and “seeking reassurance” or “not wishing to contact the GP” were graded “medium” on the grounds that other services may become more convenient, and services other than A&E may provide reassurance. Not being registered or wishing to contact a GP does not in itself preclude the future possibility of consulting providers other than A&E or a GP, so these reasons were also classified as “medium”. Reasons grouped under “lack of awareness of other services” were graded “weak” assuming that changes in consulting behaviour may be influenced by effective publicity (table 1). Eight reasons given by patients in the “other” category (box 2) were not weighted and therefore not graded formally or included in the analysis.

Results

response

Two hundred and sixty seven attenders agreed to complete a questionnaire. Of these, two questionnaires were lost and one was handed in uncompleted. Twelve of those who completed a questionnaire did not sign the consent to access the notes. Full data (a completed questionnaire and notes review) were obtained for 96% of the forms distributed (255 of 267). The achieved sample consisted of 119 from attenders triaged to the “yellow” stream and 136 from the “green” stream. The age-sex distribution of the unweighted sample was 62% men 38% women, and 57% were aged under 35 years.

estimated characteristics of attendance at a&e in a non-urgent population

After weighting of the achieved sample the estimated ratio of men to women in the population was 2:1 (68% M; 32% W). The mean age was 34 years. It is estimated that 81% were self referred, the remaining 19% having been advised to attend by someone in authority to them such as a manager, or health related personnel. The recoding of the health problems yielded an overall case mix of 80% trauma: 20% non-trauma. Slightly more than one quarter (27%) of the health problems occurred at home but almost two thirds of the study population (63%) travelled to A&E from home. In self referred patients, 24% of those delaying more than six hours had consulted (or tried to consult) another health professional for the same problem before attending A&E.

suitability of the non-urgent population for treatment by other providers

We calculated that two thirds of patients in the “yellow” stream and 54% in the “green” stream, were suitable for treatment elsewhere, yielding an overall estimate of 55% (95% CI 50%, 62%) who could be treated in at least one of the other services considered. An estimated 13% might have been directed away from A&E to selfcare by NHS Direct (95% CI 9%, 17%), and 25% by the walk in centre (95% CI 19%, 31%) (fig 1). General practice was suitable for 10% (95% CI 6%, 14%), and the MIU could have treated 38% (95% CI 32%, 44%). With the exception of those suitable for treatment only in A&E, the numbers suitable for treatment by other providers (fig 1) are not exclusive. The largest overlap in suitable service provision was between walk in centres and NHS Direct, where 28 (sample numbers) satisfied the criteria for inclusion in both (an estimated 11% of the population after weighting).

estimated differences between the population suitable for treatment only in a&e and others

The estimated age profile of those suitable for treatment only in A&E is younger than all the others. A higher proportion in the A&E group was aged under 35 years (table 2). No differences were observed between those suitable only for A&E and all the others, in the proportions that arrived at A&E by a private car, but 14 of the 18 in our sample who arrived by taxi (unweighted numbers), were in the A&E group. Of the 21 in the sample who had home addresses outside the Sheffield boundary (an estimated 9% of the population after weighting), the health problems of 13 were suitable for treatment only in A&E. Looking solely at those who self refer, the only population-based differences in the proportions of attenders at A&E assigned to low priority treatment streams between those suitable for treatment only in A&E and the others which retained statistical significance was in delaying less than six hours before presenting to A&E (table 2).

reasons for attending a&e

The reasons reported most frequently for attending A&E in our sample were: “I thought I needed a radiograph ..” (n=112), and being “...advised to come to A&E by someone else” (n=108) (table 1). The sources of advice in this second group divided between informal “friends or family” and “others”. The reason reported least often was “seeking advice from a nurse practitioner”.

In the pattern of reasons for attending A&E provided by each patient, 34 of the 255 (unweighted numbers) did not include at least one “strong” reason with a low potential to change their decision to attend A&E. In the 55% of the population who were suitable for treatment by other providers 12% (95% CI 6%, 18%) did not include at least one strong reason for attending A&E. Thus the estimated proportion of this non-urgent A&E population who might appropriately (defined according to various objective measures) and realistically (as defined by patient perceptions of need) divert to immediate care providers other than A&E in similar circumstances in the future, was just 20 of 255 (estimated 7% after weighting (95% CI 3%, 10%)).

Discussion

principal findings

Objective measures indicate that newly developing, alternative immediate care services have the potential to treat 55% (95% CI 50%, 62%) of patients currently attending A&E with less urgent conditions. However, the patterns of the patients' reasons for presenting to A&E observed in our study, suggest that the real extent to which these services will deflect demands for non-urgent care away from A&E is much less. The belief that a radiograph is necessary is clearly an important factor in the decision to present to A&E, as are disparities between the professional view and the patient's perceptions of the seriousness of the health problem, and expectations of care. Previous patterns of consulting behaviour and experiences of services both elsewhere and in A&E are also important factors. In our study just 12% of those with health problems defined objectively as suitable for treatment elsewhere, did not include a “strong” reason for presenting to A&E. It is estimated therefore that realistically, the alternative service models we considered will absorb just 7% (95% CI 3%, 10%) of the patient population currently presenting to A&E with non-urgent health problems.

validity/comparability with other studies

There is no reason to believe that our sample from two low priority treatment groups in a typical A&E department is not representative of patients in the same treatment categories seen in other A&E departments, or to doubt the wider generalisability of the results. The proportion of all A&E attenders who could be treated in primary care or minor injury setting has been estimated previously as 10%–40%¹⁷ and 40%–50%¹⁸ respectively. These proportions are consistent with our findings of 55% of the non-urgent caseload that could be treated in another setting. Our finding of 24% of self referred attenders who have consulted a health professional previously for the same health problem before presenting to A&E is similar to the proportion of 21% reported in an earlier study.¹³

mechanisms and implications for policymakers

The recent changes in primary care¹⁹ may require that the criteria we used to identify the subgroup of patients suitable for treatment in general practice be reviewed. It is nevertheless remarkable that using a validated algorithm,¹⁶ general practice was identified as suitable for managing only an estimated 10% (95% CI 6%, 14%) of the study population and was revealed as a relatively unimportant alternative to A&E for non-urgent health problems. In contrast, based on current provision in Sheffield, up to 38% of the non-urgent caseload in A&E could be treated suitably in a MIU. If the availability of the MIU was extended to 24 hours × 7 days it might be able to absorb 72% (95% CI 66%, 78%) of the non-urgent A&E caseload. We did not estimate what the costs of such changes would be but the methods used in the study would permit such comparisons to be made.

Most of the difference between the service potential of an MIU and a walk in centre to relieve the non-urgent demand for A&E is explained by access to radiography facilities. Radiography was not included in the specifications for the first 19 walk in centres declared¹⁰ but radiographs are available within strict guidelines in the model of MIU we considered. In our study, 58 radiographs were carried out in those believing a radiograph was required, and a further 37 in those not expecting to have one, indicating considerable disparity between patient expectations and professional views of treatment required. Whether or not all the radiographs ordered were clinically necessary was not answerable within our design, but “I thought I needed a radiograph ..” was the reason cited most often by our sample (table 1). The availability of radiographic services is clearly an important factor, and is likely to have an impact on the utilisation of services that are remote from radiographic facilities. On the other hand, only three patients indicated specifically that they wished to see a nurse practitioner. This again raises issues around patient expectations and the use of nurse-led services by those seeking advice from a doctor. These considerations are of great importance in attempts to encourage patients to choose appropriate services such as in the department of health's recent campaign “Choosing the right remedy”.²⁰ These campaigns are likely to falter on a confusion of clinical needs and patient perceptions and beliefs. Advice to go to A&E or MIU “if you need urgent hospital treatment” is predicated on mistaken assumptions about the capacity of patients and their friends and family to assess this accurately.

strengths and weaknesses of the study

Our confidence in the study is based on its design and rigorous execution of the methods that permitted patients' perceptions of their health problems and their reasons for attending A&E shortly after arriving in the department, to be compared directly with the actual diagnoses and processes of care received. The shortfall in recruiting 267 to complete the questionnaire rather than the target of 294 occurred in the initial stages of the study. This was attributable to misunderstandings between staff about the processes entailed rather than patients refusing to participate and was resolved by improved communication and monitoring. Potentially, selection bias could have occurred in the recruiting but we were able to confirm by cross checking the numbers assigned to the questionnaires at the point of recruitment against the sequence of numbers printed electronically on the A&E notes at the time of registration, that bias from this source was minimal.

The process of grading the strengths of the reasons given by patients for attending A&E was subjective. However, each member of the research team graded the reasons independently, and differences in scores were resolved in a panel discussion. We acknowledge the complexities in patients' consulting behaviour reported elsewhere,²¹ and recognise that interaction between different reasons is unlikely to be constant. However, the sensitivity tests we conducted around the classification of the reasons made only marginal differences to some of the results, and had no effect on the key finding, which is in accord with the results of recent evaluations of new services,^{11, 22} that observed no effect on attendances at A&E departments.

future research

A proliferation of services other than A&E where patients with non-urgent health problems might present, seems to widen patient choice and also to offer an opportunity to reduce demands on A&E. However, looking at the patterns of reasons for attending A&E provided by the patient, we found a strong impetus to continue to present to A&E. The challenge in the context of managing non-urgent A&E caseload therefore is to focus less on the clinical needs of patients and on developing services remote from A&E to meet these needs, in favour of improving our understanding of patient consulting behaviour to facilitate the appropriate matching of service to patient perceptions of need.