IA are a long-standing concern for EDs and still provide an excess burden on services . This ecological study provides insight into the demographical representation and temporal factors associated with IA to ED. The large sample has allowed for a level of analysis that has not been previously performed. Almost 12% of the attendances in our study were deemed inappropriate, a value substantially lower than others reported internationally . This marked difference is likely due to the definition of IA used. Our retrospective definition includes only attendances that were self-referred, received no investigation and either advice or no treatment, and were discharged without any follow-up or to primary care. These are easily measurable criteria likely to provide high specificity, but by using a generic definition of IA, it is unavoidable that certain individual attendances can be misallocated. For example, attendees who receive simple medication (for example, over-the-counter analgesia) that could readily be provided by other services (like pharmacy) will be grouped as appropriate, whereas attendees with certain psychiatric complaints may require emergency assessment and no other investigation or intervention, and thus be deemed inappropriate.
Reducing IA to EDs could have a significant effect on the quality and continuity of care provided to patients, and also on the overall financial cost of this service. Using our findings, IA resulted in an estimated cost of nearly £100 million between April 2011 and March 2012, assuming an ED attendance with no investigation or significant treatment cost £54 . It should be noted, however, that many costs of EDs are relatively fixed (such as staffing and opening of a department), and further research would be required to examine whether reductions in IA would result in cost saving or increased efficiency and utility of existing resources.
We found age to have a strong relationship with IA. Odds of IA were highest in the very young (peak attendances were in one and two year olds), and elevated between mid-teens and mid-twenties, followed by a steady fall as age increased thereafter. The inverse relationship between IA and age found in our study has also been identified elsewhere (for example, USA, Canada and Brazil) [20–22]. Thus, our findings suggest that interventions to prevent IA should be targeted towards early childhood and young adults in their late teens to late twenties. For young children, the decision to attend ED lies with parents and guardians, and this is likely reflective of the pressures of parenthood and a belief that the ED is the most appropriate place to receive care [23, 24]. This could be offset through targeted education to parents about the appropriate use of ED services, or by providing details of other local health services capable of providing prompt medical advice when to access to primary care and out-of-hours services available. This could be delivered routinely via health professionals who have a high degree of contact with new parents (for example, through home visits and routine health checks), such as health visitors, community midwives and nursery nurses.
The second peak in odds of IA seen from late teens to late twenties could reflect a time when young people are leaving home for the first time (such as to attend university), and may indicate a poor understanding regarding appropriate use of ED services, a lack of knowledge of other health services available, and poor access to primary care (for example, still registered with childhood general practice). Targeted education for school leavers and university students regarding appropriate use of ED, alternative health services available in the local area and the importance of primary care and registration with a local general practitioner could prove useful in reducing IA in this group.
Research in other countries has found that females are more likely to attend ED inappropriately than males (for example, Brazil, Turkey and USA [22, 25, 26]). Conversely, we found that males were slightly more likely to attend inappropriately than females, although absolute differences were small. This may reflect a difference in the definition of IA or differences in the structure and use of health services between countries. Whilst deprivation has been linked to IA in other studies, the direction of association has been mixed and may depend to some extent on the marker of deprivation used (for example, education, income, social class or postcode) [13, 25, 27]. Using a measure of residential deprivation, we found that the most deprived population accounted for the highest numbers of both AA and IA. This likely represents the poorer health and greater injury risk experienced in deprived communities. However, after controlling for age and gender, those from the least deprived quintile had the greatest odds of IA relative to AA. Several mechanisms might explain this finding, including greater access to ED among more affluent individuals (such as through increased availability of transport) , and smaller family size possibly permitting greater focus on individual children and increased concern over non-urgent conditions [23, 29]. There is a need for greater clarification around this relationship to help understand why certain social groups may be more likely to attend inappropriately than others. Although significant, the degree of difference between deprivation quintiles is only small and with IA occurring most often in the most deprived communities, measures to manage service pressures by providing additional services and addressing IA would be of greatest benefit in deprived areas.
Both AA and IA were seen to occur most regularly on Mondays, during March and between 8 am and 4 pm. When controlling for temporal effects, relative to AA odds of IA were significantly higher on weekend days, bank holidays and between the hours of 8 am and midnight. These findings can inform both the management of ED services and prevention of IA; service provision may be best targeted on Mondays and between the hours of 8 am and 4 pm, whilst measures to raise awareness may be most effective if targeted at weekends and Bank Holidays. The increase during weekends and bank holidays likely represents a lack of access to primary care during these times, and a reluctance to take time off work during the week to access these services. Although significant, the variation seen in IA by month was smaller than that seen when comparing weekdays to weekends or bank holidays, suggesting that month is unlikely to be a major factor informing service management or prevention measures regarding IA.
Internationally, numerous methods have been used to prevent inappropriate use of EDs. These include diverting calls from emergency services, ambulance non conveyance, attempts to triage out IA and general education . These interventions have experienced variable success and have raised questions over patient safety . Patient safety is paramount and any potential negative effects of intervention (for example, a delay in attendance to an ED for an urgent health problem) must be carefully considered before implementation. A further method of addressing IA, trialed in England, is the provision of primary care physicians either alongside emergency physicians in the ED itself, or attached to the department in a general practice surgery . This is intended to provide alternative options for what is deemed an IA at ED, with research suggesting it is a safe and cost-effective intervention [31, 32], and one supported by the College of Emergency Medicine . It has been suggested that primary care services are currently insufficient to manage the demand for health treatment and require modification to reduce the burden on ED .
The major strength of our study is its scope; the HES A&E dataset has provided access to a much larger sample of IA than previously studied and one that represents the majority of ED attendances in England over a one-year period. Also, the use of the Index of Multiple Deprivation has allowed for a more comprehensive review of deprivation comparative to previous literature. However, a number of limitations do exist. Firstly, alongside potential misallocation of attendances to either the IA or AA groups, only using attendances that were self-referred will have missed any inappropriate cases referred from primary care, telephone triage services or the ambulance service, while the exclusion of cases who left the ED before being treated or having refused treatment may have further missed IA. We were also unable to account for variation in staff practices regarding investigation and treatment. Despite this, our definition should act as a suitable proxy for IA, and results will remain relevant when considering prevention or management. Another limitation is the lack of additional data which is inevitable when using datasets such as the HES. Information on access to primary care services, reasons for choosing ED as point of care, general impressions of different services and patients’ own view of attendance appropriateness will be important in determining potential predictors of IA. Additionally, the incompleteness of the dataset is an important limitation. Although small relative to our sample size, over 470,000 attendances were removed because they could not be assigned to an appropriateness category. In addition, only 62.6% of attendances had a valid diagnosis code , preventing analysis of these data, which would have provided information that could further inform prevention.
This study is the first to explore IA across England as a whole using the HES dataset for ED attendances. Whilst this dataset is currently experimental, coverage across England continues to improve each year, with 80.5% of all ED attendances in England included in data for 2011 to 2012 and over 90% of cases having valid data on investigation, treatments and disposal . With an urgent need to reduce the burden on ED across England, this dataset, and the methods detailed in our study, could readily form the basis of a monitoring system, allowing for timely evaluation of interventions and services implemented to alleviate the ED burden of IA and increase the quality of the service. To strengthen the dataset, an appropriateness field could be added, which could alleviate concerns about sensitivity. To do this, a clear definition of IA would be required, based on objective criteria rather than subjective evaluation. Such a definition of IA must be highly robust and exclude any attendance with a risk of serious sequelae resulting from non-use of ED. A national policy for clinical ED staff to determine appropriateness via these criteria would allow for effective inclusion of IA into the HES A&E dataset.