Article Text

Download PDFPDF

The impact of out-of-hospital models of care on paediatric emergency department presentations
  1. Russell M Viner1,2,
  2. Frances Blackburn1,
  3. Francesca White1,
  4. Randy Mannie3,
  5. Tracy Parr1,
  6. Sara Nelson1,
  7. Claire Lemer4,
  8. Anna Riddell5,
  9. Mando Watson6,
  10. Francesca Cleugh6,
  11. Michelle Heys2,
  12. Dougal S Hargreaves1
  1. 1 Healthy London Partnerships, London, UK
  2. 2 UCL Great Ormond Street Institute of Child Health, London, UK
  3. 3 Financial Strategy, NHS England (London Region), London, UK
  4. 4 Evelina London Children’s Hospital, Guy’s and St Thomas' NHS Foundation Trust, London, UK
  5. 5 Barts Health NHS Trust, London, UK
  6. 6 Imperial College Healthcare NHS Trust, London, UK
  1. Correspondence to Professor Russell M Viner, UCL Great Ormond St., Institute of Child Health, 30 Guilford St., London WC1N 1EH, UK; r.viner{at}


Objective To estimate the potential impact of enhanced primary care and new out-of-hospital models (OOHMs) on emergency department (ED) presentations by children and young people (CYP).

Design Observational study.

Patients & setting Data collected prospectively on 3020 CYP 0–17.9 years from 6 London EDs during 14 days by 25 supernumerary clinicians. CYP with transient acute illness, exacerbation of long-term condition (LTC), complex LTC/disability and injury/trauma were considered manageable within OOHM. OOHMs assessed included nurse-led services, multispecialty community provider (MCP), primary and acute care system (PACS) plus current and enhanced primary care.

Measures Diagnosis, severity; record of investigations, management and outcome that occurred; objective assessment of clinical need and potential alternative management options/destinations.

Results Of the patients 95.6% had diagnoses appropriate for OOHM. Most presentations required assessment by a clinician with skills in assessing illness (39.6%) or injuries (30.9%). One thousand two hundred and ninety-one (42.75%) required no investigations and 1007 (33.3%) were provided only with reassurance. Of the presentations 42.2% were judged to have been totally avoidable if the family had had better health education.

Of the patients 26.1% were judged appropriate for current primary care (community pharmacy or general practice) with 31.5% appropriate for the combination of enhanced general practice and community pharmacy. Proportions suitable for new models were 14.1% for the nurse-led acute illness team, MCP 25.7%, GP federation CYP service 44.6%, comprehensive walk-in centre for CYP 64.3% and 75.5% for a PACS.

Conclusions High proportions of ED presentations by CYP could potentially be managed in new OOHMs or by enhancement of existing primary care.

  • integrated care
  • health services research
  • epidemiology
  • health service

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?

  • Improved integration of care for children and young people (CYP) is a key element of plans to improve outcomes for CYP in the UK.

  • Improvements in out-of-hospital (OOH) care for CYP have focused on either enhancing existing primary care or on new integrated models of care.

  • The impact of introducing new models of care is difficult to predict without detailed clinical studies assessing the potential for management in new models.

What this study adds?

  • Just over a quarter of emergency department (ED) presentations could appropriately be managed within the current primary care offer, that is, either in community pharmacy or general practice.

  • New OOH models offer the potential to manage from 14% (in limited nurse-led services) through to 75% (in a primary and acute care system model) of current ED presentations.

  • Further work is needed to understand the potential economic costs and benefits of new models, including data on family choice and behaviour.


Improved integration of care for children and young people (CYP) across levels of the health service and across sectors including education and social care is a key element of plans to improve outcomes for CYP in the UK.1–3 A major component of better integrated care is reducing unnecessary hospital attendances and providing better quality care closer to home when possible1 4 5 consistent with national plans to expand ’out of hospital' (OOH) care.6 Rising emergency department (ED) attendances among CYP, 20% higher than a decade previously,7 have brought the need for improved OOH models (OOHMs) into very sharp focus for those who commission and provide CYP health services.

Improvements in OOH care for CYP have focused on either enhancing existing primary care or on new models of care. Enhancement of primary care has included improving access by extending hours in general practice and community pharmacies, given evidence that families generally prefer to access general practice for acute care if services are open and accessible,8 that CYP with better access to general practioners (GPs) have fewer ED visits9 and that most of the population live within walking distance of a pharmacy.10 A further innovation has been the development of GP federations to provide benefits at scale including improved access, a greater emphasis on health promotion and the potential for a more specialist offer for CYP.11

A range of new OOHMs of care have been developed and evaluated for adults in the UK, using new funding, accountability and workforce models.6 However, progress in developing models of care for CYP has lagged behind,5 despite CYP aged 0–18 years comprising 24.9% of ED presentations in England in 2015–2016.12 Although some innovative CYP-focused pilot models exist across the country,13 14 there are few data available to support more widespread implementation of new acute models of care for CYP, despite commissioners actively examining how new models could be commissioned.5

Although population-level variation in levels of OOH care for CYP are well described,15 the impact of introducing new models of care is difficult to predict without reviewing the clinical needs of CYP attending ED and assessing the necessary level of care in each case. To estimate the potential impact of enhanced primary care and new OOHMs on CYP ED presentations, we used data on clinical need from a prospective clinical cohort presenting to EDs across London to identify proportions of CYP that could be appropriately managed within each OOHM, thus reducing ED presentations.


Design: observational study.


We purposely recruited six sites which had not implemented significant integrated care programmes for CYP, including sites from both north and south London, from boroughs with deprivation affecting children from among the lowest to among the highest in in England and boroughs with Accident & Emergency (A&E) presentation rate per 1000 children aged 1–4 years similar to England average. Sites were St George’s Hospital, St Helier Hospital, Kingston Hospital, the Royal London Hospital, Newham Hospital and Whipps Cross Hospital. For site characteristics, see appendix table 1 supplementary file 1.

Supplementary file 1

Questionnaire used to collect data

Sample size

ED attendance data suggested that 14 days of data collection should provide a sample of 3–4000; a sample of 3000 would provide precision of ±1.5% for a proportion of 20% eligible for any OOHM. The precision estimate was obtained using the cii proportions command in Stata V.14 (StatCorp, College Station, Texas, USA).


We prospectively collected data on all CYP registered as attending participating EDs during 14 consecutive winter days (22 February to 06 March 2016), limiting data to those attending from 10:00 to 22:00 as this encompassed the peak attendance period in all trusts and the likely opening hours of new OOHM.

Eligible CYP were(1) those streamed by the hospital to attend the ED (rather than any other part of the hospital, for example, urgent care centre (UCC)) and (2) who were given a clear diagnosis or management plan by ED/paediatric staff between 10:00 and 22:00.

We excluded CYP who were not directly managed in the ED, that is, those managed within an onsite GP-led UCC and those referred directly from their GP to a paediatric assessment unit (PAU) or acute short-stay unit. Age criteria varied by hospital site, either <16 years or <18 years, reflecting national variation in service delivery for children in ED.

CYP attending ED were divided into seven segments by diagnosis: transient acute illness, exacerbation of long-term condition (LTC), complex LTC/disability and injury/trauma, safeguarding, mental health and non-trauma surgery. While all CYP were included in the study, only the first four segments were considered potentially manageable within OOHM.

Data collection

Non-patient-identifiable data were collected by a team of 25 supernumerary study clinicians (from general practice, emergency medicine, paediatrics; each with ≥4 years postgraduate experience) who had no clinical responsibilities during the shift. Study clinicians were provided with an hour’s training by telephone using standardised written information. These study clinicians obtained data from the ED staff managing each of the patients in real time, and recorded data in an online secure data collection system.

Data collected (see appendix table 2) supplementary file 2:

  • Details of presentation: time of presentation, date of birth, patient segment, diagnosis and severity.

  • Management given in ED: investigations and management received, staff types and seniority involved, and outcome destination.

  • Objective assessment of clinical need: Clinical needs and most appropriate skill set and timeframe with which to address these needs. Options for most appropriate clinical skill set were one of1 pharmacist only;2 illness clinician who is competent in assessing an acutely unwell CYP, and has immediate telephone or face-to-face access to a paediatric consultant if required;3 clinician competent in assessing an acute injury;4 specialist paediatrician; or5 other specialist. Timeframe options for each were within 4 hours, 12 hours or 48 hours.

  • Potential alternative destinations for each child given the existence of OOHMs providing follow-up or observation facilities.

Supplementary file 2

Questionnaire used to collect data


OOHMs assessed included those previously identified by us that aimed to prevent ED attendance16 and models taken from the8  National Health Service (NHS) England 2015 Five Year Forward View,6 such as multispecialty community provider or primary and acute care system (PACS) models. For comparison, we also evaluated proportions of CYP that could be managed in current primary care (eg, general practice and community pharmacy), and in future enhanced primary care models including enhanced general practice and a GP federation model. Models are summarised in table 1. Algorithms were derived a priori to assign CYP as eligible/not for each model, based (in order) on segment, diagnostic group, severity, age range, opening hours and days of service, clinician expertise required and timescale, investigations needed and timescale and management needed and timescale (see table 1).

Table 1

Out-of-hospital (OOH) models of care assessed in this study


Data were cleaned by hand, including assigning free-text data on diagnosis, investigations or management to existing or new categories and ensure consistent assignment of diagnoses to segments. Algorithms were then used to assign CYP as appropriate or not for each OOHM. Note that CYP could be appropriately managed in multiple models of care. We described characteristics of the overall sample and the sample for each model. Assessment of effects of site and observer were made using multilevel models including random effects for site and study clinician. Analyses were undertaken in Stata V.14.


This work was identified as service evaluation not requiring patient consent by the National Research Ethics system (NRES). Caldicott Guardian permissions were obtained at each site.


Data were collected on 3021 patient episodes with 1 excluded as ineligible due to age, making the sample for these analyses 3020. The 25 study clinicians entered data on an average of 121 (range 33 to 569) patients each. Patient characteristics are shown in table 2 and were similar across sites. Of the patients 95.6% were within the four segments appropriate for OOHMs.

Table 2

Patient characteristics for the overall sample (all segments) and by hospital site

Assessment, investigation and management needs are shown in table 3. The majority of presentations required assessment by a clinician with skills in assessing illness (39.6%) or injuries (30.9%). One thousand two hundred and ninety-one (42.75%) required no investigations and 1007 (33.3%) were provided with no treatment other than reassurance. Table 4 shows patient destination together with the potential alternative destination judged appropriate for clinical need. Of the presentations 42.2% were judged to have been totally avoidable if the family had better health education.

Table 3

Patient needs for assessment, investigation and management for overall sample (all segments) and by hospital site

Table 4

Destinations and potential alternative destinations for whole sample and by hospital site

Table 5 shows the proportions of all CYP likely to be appropriately managed within each OOHM of care. 26.1% were appropriate for current primary care (community pharmacy or general practice). Proportions suitable for each new OOHM ranged from 14.1% for a nurse-led acute illness team through to 75.5% for a PACS model. The combination of enhanced general practice and community pharmacy was appropriate for 31.5% of presentations.

Table 5

Proportions of CYP potentially manageable within current and new OOH models of care

There were very marked age differences for primary care models and for enhanced illness models, with proportions of children <5 years appropriate for these models being nearly double that of older children. There were notable differences across site in proportions appropriate for primary care and illness models but not for comprehensive models.

In multilevel models for each OOH, intraclass correlation coefficients (see appendix table 3 supplementary file 3) for site and for clinician observer were low, indicating little variation at the site or observer levels, with the exception of community pharmacy.

Supplementary file 3

Intraclass correlation coefficients (ICCs) for site and observer obtained from multilevel models for the likelihood of being appropriate for each OOH model, adjustedfor age


Assessments of clinical needs in real time by experienced clinicians in a large sample of CYP presenting to London EDs suggests that high proportions of ED presentations by CYP could potentially be managed in current or new OOHMs. Just over a fifth (22.3%) could have been appropriately managed in current general practice, with a quarter (26%) manageable within the current primary care offer, that is, either in community pharmacy or general practice. Enhancements to the general practice offer including extended opening hours and ready access to senior paediatric advice increased this to 28% for general practice alone and to 32% when enhanced general practice was combined with current models for community pharmacy.

Further enhancement of the primary care offer for CYP across a GP federation, including access to experienced paediatric nursing and capacity to treat minor injuries as well as rapid access to senior paediatric opinion, expanded this to nearly half (44.6%) of ED presentations. New nurse-led models in primary care or the community also present considerable potential to reduce ED presentations (eg, 28.4%) if they provide walk-in opportunities and the range of conditions treated is not excessively narrow. However services focused on ill children with a restricted range of conditions have only a limited potential to reduce ED presentations. More comprehensive models offer the potential to manage very large proportions of ED presentations, up to 75% for a PACS model for CYP. The potential for new models to reduce ED presentations was markedly higher for children <5 years for primary care and illness-only models, reflecting the higher proportions of injuries or serious presentations seen among older CYP in our data.

There were differences across sites in proportions appropriate for different models, particularly for primary care and illness-only models. However, our multilevel analyses suggested that variation largely lies at the individual rather than the site level, that is, that the apparent differences between site represent differences in the types of individuals seen at the sites rather than an inherent difference in how the sites function. We speculate that these variations are likely to reflect differences in the populations presenting to EDs, with some seeing a larger proportion of younger children, particularly for febrile illnesses, as non-use of primary care for febrile children is higher in deprived communities.17

We are not aware of similar previously published studies of the potential impact of various OOHMs on ED presentations by CYP. Our study clinicians judged that around 26% of ED presentations could be managed within current primary care and that 50% could have been avoided with better health promotion and/or greater family confidence in self-management. These are consistent with previous systematic review findings that 20%–24% of all-age ED presentations were inappropriate,18 highest among CYP,19 and findings that low health literacy is associated with higher risk of ED presentation.20 Our finding that enhancement of the primary care offer in terms of improved patient access and availability of senior paediatric advice increased the proportion that could be managed within primary care is consistent with evidence that CYP in England attending GP practices with higher levels of access have lower ED presentations9 21 and short-stay admissions,9 and that routine availability of senior paediatric advice in primary care reduces ED attendances.13

The chief limitation of our study was that we could not address the very important issues relating to patient choice or health-seeking behaviour or the practicalities of funding and staffing alternative models. Parental attendances to EDs may not be easily diverted in reality despite the existence of alternative services, with habit, issues relating to knowledge and the assumption of expertise residing in hospitals likely to strongly influence family behaviour. Some of the models examined here do not currently exist, and the necessary workforce (eg, advanced paediatric nurse practitioners, additional GPs and community nurses) to staff them or existing models are not yet available. Systems to stream CYP to attend community or hospital services are not yet available, although NHS 111 and similar services may ultimately provide this. Our data cannot address these issues, but instead provide an estimate of the likely maximum potential for new models to reduce ED attendances, that is, if all parents chose to and could attend the appropriate service and adequate staffing were available to run the services.

Our study had a number of other limitations. We obtained data on 52% of those recorded as presenting to ED at our sites during the study period, however the true denominator for our study is lower than this as those treated in PAUs or UCCs may have been registered with ED but may not be not eligible for our study. We could not examine whether the proportion of attendances included differed by time or day at each site, and the direction of any biases from this is unclear. Estimates of proportions suitable for each model are based on clinical skills, investigations and management options we judged were available in each model. These are not standardised and our data cannot account for variations in provision within models. We did not collect data before 10:00 so we could not fully assess models with earlier opening hours. Our data apply to winter, when acute illness presentations are higher than in summer months. We were unable to collect identifiable data including gender and ethnicity or data on deprivation; however these were not directly relevant to the aims of the study. We could not track which of the episodes within our data set were reattendances within the study period. However, this is unlikely to be an issue as unplanned reattendance rates within 7 days are <5%.22


Our data provide support for the enhancement of current primary care and commissioning of new OOHMs to reduce ED presentations and improve quality among CYP, particularly children <5 years. Enhancement of access to primary care in England is already a high priority for NHS, and our data suggest that modest additions to the primary care offer offers significant potential to reduce ED presentations, if appropriate incentives are in place for families to use these services and sufficient primary care capacity is available. More comprehensive models, such as a PACS service for CYP, offer the greatest potential to reduce ED presentation, yet we are not aware of current PACS models operating for CYP in the UK. Further work is needed to understand the potential economic costs and benefits of new models, including data on family choice and behaviour. The models studied here are not exhaustive, and these data could be used to study additional future models.


The authors thank the Clinical Advisory Group for this project, namely: Dr Eugenia Lee, Dr Oliver Anglin, Satish Bangalore, Jeanette Barnes, Prof Mitch Blair, Dr Hillary Cass, Ruth Chapman, Dr Ronny Cheung, Dr Graeme Hadley, Georgie Herskovits, Atefe Hossain, Michelle Johnson, Dr Mirtuza Khan, Dr Robert Klaber, Dr Monica Lakhanpaul, Annette Langseth, Kay Larkin, Dr Ian Maconochie, Donal Markey, Dr David Masters, Louise Morton, Dr Tonia Myers, Sara Nelson, Dr Trisha Radia, Dr Meredith Robertson, Dr Andrew Robins, Dr Sabah Salman, Dr Jenny Selway, Sarah Shade, Dr Gurpreet Singh, Fiona White and Dr Ingrid Wolfe for their valuable inputs. The authors thank the study clinicians who undertook the data collection, namely Drs Andras Meczner, Abigail Whitehouse, Dinesha Mapa, Libby Keating, Ghada Saleh, Joanna Eyeson, Jocelyn Guttery, Jenna Haddad, Jay Krishnan, Jackie Pitchforth, Joe Ward, Logan Manikam, Lisa Van Geyzel, Maanasa Polubothnu, Mark Ratnarajah, Poppy Flanagan, Prita Rughani, Anjum Bahadur, Sarita Gorolay, Sonia Tsukagoshi and Umiara Saleem. The authors also thank the study clinicians who undertook the data collection: The authors also thank the staff of the local hospitals (Clinicians, Staff Bank, Coding and Business Intelligence Unit analysts, Finance Leads), who made this study possible.



  • Contributors The study was conceived and designed by RMV, TP, FB and RM. CL, AR, MW, FC and DSH helped to refine the design. FB, FW, SN and TP organised and undertook data collection. RMV led the analyses, with input from FB and FW. RMV led the writing of the paper, with input from DSH, FW, TP and MH. All authors approved the final version of the paper.

  • Funding Funding for this study was provided by the Children and Young People’s Board of Healthy London Partnerships, a transformation programme jointly funded and administered by NHS England (London Region) and the London Board of CCGs.

  • Competing interests None declared.

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

  • Data sharing statement Data are available upon request to the first author, RMV.

Linked Articles