Article Text

Download PDFPDF

Preschool children who are frequent attenders in emergency departments: an observational study of associated demographics and clinical characteristics
  1. Mitch Blair1,
  2. Alan J Poots2,
  3. Valencia Lim2,
  4. Stephen Hiles3,
  5. Geva Greenfield2,
  6. Caroline Crehan4,
  7. Beverley Kugler4,
  8. Charlotte Boreham4
  1. 1 Department of Paediatrics, Imperial College London, London, UK
  2. 2 NIHR CLAHRC Northwest London, Imperial College London, London, UK
  3. 3 Department of R and D, Northwick Park Hospital, Harrow, UK
  4. 4 Department of Paediatrics, Northwick Park Hospital, Harrow, UK
  1. Correspondence to Professor Mitch Blair, Department of Paediatrics, River Island Academic Centre for Paediatrics and Child Health, Imperial College (Northwick Park Hospital campus), Harrow, Middlesex HA1 3UJ, UK; m.blair{at}


Background Unscheduled visits to emergency departments (ED) have increased in the UK in recent years. Children who are repeat attenders are relatively understudied.

Aims To describe the sociodemographic and clinical characteristics of preschoolers who attend ED a large District General Hospital.

Method/study design Observational study analysing routinely collected ED operational data. Children attending four or more visits per year were considered as ‘frequent attenders’. Poisson regression was used with demographic details (age, sex, ethnicity, sociodemographic status) to predict number of attendances seen in the year. We further analysed detailed diagnostic characteristics of a random sample of 10% of attendees.

Main findings 10 169 patients visited in the 12-month period with 16 603 attendances. 655 individuals attended on 3335 occasions. 6.4% of this population accounted for 20.1% of total visits. In the 10% sample, there were 304 attendances, and 69 (23%) had an underlying chronic long-standing illness (CLSI). This group were 2.4 times more likely to be admitted as inpatients compared with those without such conditions, median length of stay of 6.2 hours versus 2.5 hours (p=NS).

Conclusions Frequent ED attenders fall broadly into two distinct clinical groups: those who habitually return with self-limiting conditions and those with or without exacerbation of underlying CLSI. Both groups may be amenable to both additional nursing and other forms of community support to enhance self-care and continuity of care. Further research is required to increase our understanding of specific individual family and health system factors that predict repeat attendance in this age group.

  • Accident & Emergency
  • General Paediatrics
  • Health services research
  • Data Collection
  • Epidemiology

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • Frequent attendance in emergency departments is well described in adult medicine but less so in paediatrics.

  • Primary care sensitive conditions should be managed outside hospital wherever possible.

What this study adds?

  • Six per cent of under-fives were frequent attenders (four or more times per year), accounted for 20% of the workload in the emergency departments.

  • Approximately, one-quarter of these children had a chronic long-standing illness (CLSI) with the remainder being frequent attenders with self-limiting illness with no record of CLSI.

  • The majority of repeat attenders stay for less than 6 hours.


Unscheduled visits to hospital-based emergency and urgent care services in England have increased by 25% in the past decade.1 There has been a rise of recurrent and unscheduled visits which in themselves contribute a disproportionate amount to total emergency visits worldwide.2–7 This group of recurrent attenders, sometimes referred to as repeat or frequent attenders account for approximately 8% of total emergency department (ED) attendances in the UK.8 The individuals are a heterogeneous group, often with complex health conditions, multiple comorbidities and social and financial problems.9 10 In addition, there are health system factors such as dissatisfaction or difficulties with access to primary care and variable patterns of primary care referral.11 12 There is a relative paucity of detailed research of paediatric repeat attenders (PRAs).

In 2001, Armon et al reported that 83% of all paediatric attendances are for one of the six common presenting problems: breathing difficulty, febrile illness, diarrhoea and or vomiting, abdominal pain, seizure and rash.13 In 2011, Sands et al reported that over time, the type of presenting problems in the same ED remained almost identical, with a large proportion judged by clinicians to be primary care sensitive conditions.14 PRAs differ between different hospitals in the UK, suggesting there may be cultural, sociodemographic or health system factors accounting for these differences.15 This study aimed to establish which sociodemographic or clinical characteristics account for the increased likelihood of recurrent attendances to ED.

Specifically, we hypothesised that in the population served by our hospital, PRAs in the under-fives were more likely to be socially disadvantaged, from minority ethnic groups and male. In addition, we hypothesised that PRAs would fall into two broad groups: those with chronic long-standing illness (CLSI) and those who decide to attend out of preference with other sources of support, which we refer to as ‘mostly well or non-chronic’ (NC).


Setting, participants and study design

The hospital serves two neighbouring boroughs in North West London: Brent and Harrow. They are characterised by under-fives ED attendances of 824.4 per 1000 and 768.9 per 1000, respectively, significantly greater than the England average crude rate of 525.6 per 1000.16 To capture all unscheduled care attendances on site, we also included attendees at the Urgent Care Centre (UCC) co-located in our hospital. Our unit has a four bed short-stay paediatric assessment unit (PAU) and 26 inpatient ward beds. The detailed functioning and evaluation of the PAU has been described previously.17 A retrospective observational study using routine operational data collected within the co-located UCC and the Accident and Emergency (A&E) department (collectively the ED) in a large District General Hospital. Data were analysed for all children under the age of five attending between 01 October 2013 and 30 September 2014 who were resident during this time in either borough served by the hospital (Brent and Harrow). Data were pseudonymised to preserve confidentiality of patients and allow researchers to match multiple visits by the same patient. Referrals on the same day between UCC and A&E were removed as this represented an administrative change within a single visit, thereby preventing treating the same spell of care as two independent observations, when they actually belong to the same spell of care. Ethics approval (proportionate review) was obtained from London North West Healthcare Trust Research Ethics Committee. Ethnicity was reclassified from two distinct coding systems into standard Office for National Statistics (ONS) codes. The index for multiple deprivation (IMD) was assigned by matching with the lower super output areas (LSOAs). From an initial extract of 18 193 attendances, 16 603 with valid data entered the analytic data set equating to 10 169 individual patients.

There are various definitions used for repeat attenders ranging from 3 to 12 unscheduled visits within a period of 12 months, figures that are chosen either arbitrarily or based on a proportion of total ED visits.3 5 7 9 Clinical characteristics were extracted from a randomly selected 10% sample (using Excel random number generation) of patients who attended four or more times. This cut-off was selected on purely pragmatic grounds. Three clinicians (CC, BK, CB) characterised the conditions based on the free text and diagnostic categories recorded on the various hospital information systems. A common International Classification of Diseases-10 (ICD-10) coding framework was used for consistency. Each patient record was scrutinised to determine documented CLSI. Patients with no evidence of CLSI in any of their attendance documentation were labelled NC. Any differences of classification were agreed on by consensus. For example, mild eczema and atopy was not considered a chronic condition whereas serious eczema requiring dermatology or paediatric outpatient follow-up was considered chronic.


Descriptive analyses were performed to determine the overall demographic profile of the study population differences between PRAs and non-PRAs. The study population demography was compared with that of the general population using univariate χ2. A Poisson regression using demographic details (age, sex, ethnicity, IMD) was used to predict number of attendances seen in the year. Age was treated as ordinal, as there are only five possible values in this data set; this means we can obtain direct ORs for each age. Because the predictor variables in which we are interested are those at first presentation (ie, what was the initial age, deprivation, etc), we append the maximum number of attendances noted in 1 year to the index visit, removing repeated observations of the same individual to ensure independence of observation when running the Poisson model. Admission risk, length of stay in the department and intervisit median periods were compared and analysed for both groups (CSLI and NC) of children using Mann-Whitney U non-parametric tests. The clinical data was analysed by diagnosis for each attendance and patients with CLSI and NC compared in terms of differences in clinical profiles.


Re-attendance rates

Single visit patients (6539) account for the majority (64.3%) of the study population (10 169 patients; 16 603 attendances) and largest fraction of ED visits (39.4%). Using four or more visits as a threshold, there are 655 PRAs, with a total of 3335 attendances over the year; thus 6.4% of this population accounts for 20.1% of total visits. The highest number of recurrent visits is 16, represented by one patient (table 1).

Table 1

Distribution of study population by number of visits per patient

Sociodemographic characteristics

The largest group of attendees among all under-fives attending were male infants of Asian or Asian British ethnic origin. Most attendances at the urgent care centre occurred during working hours with higher rates on the weekend (table 2). These are statistically significant in relation to the general background population. Younger aged children from a non-Asian ethnic group who were from more socially disadvantaged backgrounds were significantly more likely to be recurrent attenders than the study population for full Poisson analysis. Online supplemental table for full Poisson analysis.

Supplementary Material

Supplementary Table
Table 2

Demographic characteristics of study population (n=10 169), compared with the under-5-year-old population of Brent and Harrow and p-value of difference by χ2

Clinical characteristics

In the 10% sample (65/655) of recurrent attenders, there were 304 attendances, 60 of these were re-attenders within 3 days. Seventy-seven per cent of the sample are children with no underlying recorded chronic or long-standing illness (NC). The remaining 23% had some form of underlying chronic illness (CLSI). A significantly high proportion of attendance in patients with CLSI resulted in PAU or inpatient admission (χ2=9.7, p=0.003). The relative risk of being admitted in the CLSI group compared with the NC group was 2.36 (95% CI 1.38 to 4.04). Mean length of stay was short in both groups with no significant difference in medians (table 3).

Table 3

Admission risk and length of stay (LOS) in patients with chronic long-standing illness (CLSI) versus no underlying chronic illness (NC)

Upper respiratory tract infections, viral illness with or without fever and rash were the most common diagnoses (table 4).

Table 4

Top diagnoses by patient group and attendance outcome (n=304 attendances)

Table 5 indicates the clinical and attendance characteristics (age and frequency) of the 15 children attending with underlying CLSI. One-third (5/15) had conditions arising in the neonatal period.

Table 5

Clinical and attendance characteristics of 15 children classified as chronic long-standing illness


In this study, 6% of under-fives were frequent attenders (four or more times per year), accounting for 20% of the workload in the EDs. Approximately, one-quarter of these children had a CLSI with the remainder being frequent attenders for unknown reasons. Similar to recent research by Wijlaars et al, we found that those with CLSI were more than twice as likely to be admitted to the hospital than those without although in our study, length of inpatient stay was not significantly different.18

Continuity both in primary and secondary care is likely to be an important factor in helping to contain minor illnesses within the community. Neill et al have shown that families vary considerably in their ability to contain such illness within the home, related to previous family experience and confidence in dealing with acute illness.19 Parental trust in medical advice is enhanced by ‘safety netting’—a technique by which medical and nursing advice supports parental decision-making if and when to seek further medical attention.20 This may be an important factor here in preventing recurrent unscheduled attendance. In our 10% sample (n=65), the number of children who returned to the ED at least once, within a 3-day period in an unscheduled event was 39 (60%), suggesting either that the illness progressed in an unexpected way that safety netting advice was omitted or inadequate or that parental resources in dealing with the illness were exceeded;

Although Asian children seem over-represented in ED attendances generally, the explanation for the significantly lower PRA rate in Asian children compared with the non-Asian or unstated ethnic groups is of some interest and may reflect a number of possible explanatory factors, among these, greater levels of ‘illness containment’,19 a differential illness rate, alternative hospital use or more effective safety netting. In our population, this difference was clearly not related to levels of deprivation as measured by IMD. This would be worthy of replication and further detailed health service usage and clinical data analysis in other localities.


To the authors’ knowledge, this study is the first published analysis of children’s ED attendance data from both A&E and UCC settings within a geographical area. Furthermore, this study contributes to the literature on recurrent paediatric attendance. It provides insight on the demographic and clinical characteristics of paediatric patients under the age of five, an important age group in emergency attendances that has been under researched. The insights from this study have indicated fruitful areas of further research around parental decision-making around acute mild to moderate illness, description of patient flow across the health system,21–23 measuring and improving continuity of care for children with CLSIs and improving signposting and access to other primary care and community resources.


There are limitations within this study. Missing values in the ED records may have led to exclusion of attendances and information that may have been significant to the research questions. Differing definitions have made it challenging to compare results with existing literature. It is difficult to know what justifies a generalizable definition of recurrent paediatric attendance. More needs to be done to explore and determine a suitable definition, especially in relation to the clinical significance of PRAs.

The findings of this study were derived from the database of a single hospital in London and may not be representative or generalisable to the population of London.

This study was limited to analysing only demographic and clinical characteristics without taking into account other behavioural, cultural and health system determinants that may impact the likelihood of recurrent attendances. Examples of such determinants include health-seeking behaviours of carers, healthcare utilisation of carers, beliefs and perceptions of ED services and even the ability to obtain a timely GP consultation. These factors are unable to be considered in this study as these data were not available but it should warrant further research to identify health disparities among PRAs and reduce attendances.

Using area measures of deprivation is a proxy for deprivation levels of individual families and patients and therefore prone to ecological bias.24 Caution is required in generalising ethnic group risks as we were unable to ascertain whether the same individuals utilised other urgent and emergency care facilities in different neighbouring localities or indeed their own GPs. This emphasises the point that such health services research needs to take into account a ‘whole system’ approach at a time where there is so much choice of provider and organisational churn in the NHS.


Jake Clements for additional statistical support. Jamie Reeves for administrative help in preparation of the manuscript.



  • Contributors MB and AJP conceived the study. MB wrote the final draft. SH extracted data and provided interpretation. AJP and GG carried out statistical analysis. VL wrote the socio-demographic section as part of her Masters thesis. CC, BK and CB contributed to drafts and conducted analysis of clinical records.

  • Funding This study presents independent research commissioned and funded by the National Institute for Health Research (NIHR) under the Collaborations for Leadership in Applied Health Research and Care (CLAHRC) programme for North West London (application reference CLAHRC-2013-10012). The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

  • Competing interests None declared.

  • Ethics approval London North West Healthcare Trust R and D committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Unpublished data has not been made available to others outside the research team.