Article Text

Download PDFPDF

Readmission in children's emergency care: an analysis of hospital episode statistics
  1. Peter Callery1,
  2. Richard G Kyle1,
  3. Malcolm Campbell1,
  4. Michele Banks1,
  5. Susan Kirk1,
  6. Peter Powell2
  1. 1University of Manchester, Manchester, UK
  2. 2Royal Bolton Hospital, Farnworth, Bolton, UK
  1. Correspondence to Professor Peter Callery, Chair in Children's Nursing, University of Manchester, University Place, Oxford Road, Manchester M13 9PL, UK; peter.callery{at}manchester.ac.uk

Abstract

Aim To compare rates of emergency readmission following discharge for common paediatric conditions from a range of hospital services.

Design Retrospective analysis of hospital episode statistics (HES) and telephone survey of service provision.

Setting Twelve hospitals serving a metropolitan area in the North West of England.

Outcome measures Emergency admissions to hospital within 7 days of discharge for breathing difficulty, feverish illness and/or diarrhoea.

Results HES were obtained for all children under 15 years of age discharged following emergency admission for breathing difficulty, feverish illness and/or diarrhoea during 2005/2006 (n=20 354) or 2006/2007 (n=23 018). The readmission rate for all hospitals in 2006/2007 was 5.5%. The percentage of same day discharges was associated with readmission (Kendall's taub correlation=0.61, p=0.007). Readmissions were also associated with the proportion of same day discharges for breathing difficulty (Kendall's taub=0.83, p<0.001) and feverish illness (Kendall's taub=0.50, p=0.023) but not significantly so with diarrhoea (Kendall's taub=0.37, p=0.098). The total number of admissions at a hospital in the year was associated with its readmission rate (Kendall's taub=0.71, p=0.002). Most of the sample lived in the 40% most deprived areas in England, but there was no significant association between readmission and living in the 10% most deprived areas.

Conclusions Readmission rates are associated with higher numbers of annual admissions and higher proportions of children discharged on the day of admission. Variations between hospitals suggest that other factors can also affect readmission rates. Readmission rates calculated from HES can contribute to assessments of the outcome of emergency services.

Statistics from Altmetric.com

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.

Introduction

Emergency hospital admissions of children in England increased by 18% in the 10 years to 2007 and the proportion of children discharged on the day of admission or the next day increased from 59% to 71%.1 The increases in same day discharge coincided with the introduction of observation and assessment units (OAU) for more extended observation without overnight hospital stay2 so that children who do not require complex interventions can avoid unnecessary hospitalisations.3 A review of patient records in the 1990s found that half of an English hospital's emergency admissions were considered inappropriate by paediatricians,4 and more recently 30% of admissions and 56% of days of hospitalisation were categorised as inappropriate in an Italian hospital.5 Guidelines for triage,6 for specific conditions such as bronchiolitis,7 and for audit and feedback8 have all been proposed to prevent avoidable hospitalisations.

What is already known on this topic

  • Emergency admissions increased by 18% in England from 1996/97 to 2006/07 but the average length of admission halved in the same period.

  • Observation and assessment facilities have been introduced to avoid inappropriate overnight hospital stays.

  • Readmission rates have not been measured consistently to enable examination of the effect of shorter hospitalisations.

What this study adds

  • Readmission rates calculated from hospital episodes statistics can be used in comparisons of emergency services for children.

  • Higher readmission rates are related to higher proportions of children discharged on the day of admission and higher numbers of discharges.

  • Variations between hospitals suggest that other factors can also affect readmission rates.

A systematic review of 25 studies of hospital based alternatives to acute paediatric medical admission suggested that OAUs can be safe, efficient and acceptable alternatives to overnight admission,9 although the limited quality and quantity of evidence suggested the need for more population based research.9 Evaluation of children's emergency care is hampered by the lack of measurable outcomes and the rarity of serious adverse outcomes.10 Readmission rates are potentially an important indicator of risk associated with early discharge. While some readmissions are inevitable and may be desirable, increased rates could indicate that children's problems have not been appropriately assessed and/or treated, or that a longer period of observation was required for an underlying condition to become apparent.

Readmissions can be calculated using hospital episode statistics (HES), a record of all admissions to National Health Service (NHS) hospitals in England.11 The introduction of the Payment by Results system increased the financial incentives for NHS organisations to record hospital admissions comprehensively.12 Nevertheless, there are coding errors in HES, which are more frequent in complex activities (including cancer treatments), multiple diagnoses (including care of older people) and poorly documented care (including minor procedures in rheumatology outpatients).13 Error rates are higher with more specific diagnosis codes13 and by region (North West England has one of the lowest error rates).13 However, common presentations for emergency medical care can be captured at the ‘block’ level of the ICD-10 diagnosis coding structure with the addition of selected specific codes.14

Previous studies have reported children's readmission rates ranging from 0.4% to 7% across durations of 2, 3, 7, 14, 21, 28 and unspecified days.8 9 15 The risk of capturing unrelated admissions could be expected to increase with longer time periods. While any time period is arbitrary in the absence of information about circumstances, a week may be a reasonable time to expect that readmission is not required if a child's condition has been appropriately managed.

Methods

The aim was to compare rates of emergency readmission of children within 7 and 28 days following emergency admission for common presentations in paediatrics: breathing difficulty, feverish illness and/or diarrhoea.

The design incorporated a telephone survey of service provision and a retrospective analysis of HES.

Telephone interviews were conducted with key contacts at 12 hospitals serving a metropolitan area in the North West of England to describe emergency services for children and specifically whether an observation and assessment service was provided for children to be discharged within 6 h. Informed consent was obtained from interviewees. The study was assessed as not requiring ethics approval by a NHS Research Ethics Committee.

HES were obtained for all children less than 15 years old discharged during financial years 2005/2006 or 2006/2007 following emergency admission to any of the 12 hospitals. An extract from HES was obtained and processed to contain details of emergency admissions for diagnoses associated with the three commonest reasons for A&E presentation: breathing difficulty, feverish illness and/or diarrhoea.16 ICD-10 codes and further details about the extract specification and data processing procedures are outlined in the online ‘Methods’ supplementary file.

Emergency readmissions within 7 days were identified as those occurring between 0 and 6 days (inclusive) of discharge. Same day emergency readmissions were included. Transfers were excluded from the numerator. Due to the likelihood of co-morbidity or the presence of an undiagnosed condition at the time of discharge, all readmissions within 7 days for children with any of the specified diagnoses were included in the numerator. Denominator data comprised the number of live discharges for each hospital provider.

Readmission rates were calculated as the percentage of live discharges from a particular hospital that resulted in readmission to any other hospital in England within 7 days. Readmission rates were then disaggregated by duration (ie, short spells lasting 0 or 1 day and longer spells lasting 2, 3 or 4 days) and diagnosis category of the discharge (ie, breathing difficulty, diarrhoea, feverish illness).

The unit of analysis was the hospital of discharge. Data were analysed descriptively using SPSS Release 15, with associations measured using Kendall's taub correlation due to expectedly skewed distributions.

Results

The number of live discharges for breathing difficulty, diarrhoea and feverish illness was 20 354 (16 841 children) in 2005/2006 and 23 018 (18 628 children) in 2006/2007. Seven hospitals reported that they operated a dedicated children's emergency OAU either within the Accident and Emergency (A&E) department or based on a children's ward. The remaining five offered children's acute services with paediatrician assessment either via A&E or open access to children's wards. Two were specialist children's hospitals and 10 district general paediatric units; bed numbers ranged from 22 to 143 (table 1).

Table 1

2006/07 Emergency readmission rates within 7 and 28 days

The proportion of children discharged on the day of admission or at day 1 (ie, after midnight following admission) ranged from 56.5% to 85.1%. Discharge on days 2 and 3 ranged from 11.1% to 31.6%. Hospitalisations of 4 or more days ranged from 3.8% to 13.4% (table 2).

Table 2

Hospitals ranked by percentage of total live discharges in 2006/07 with short (0 or 1 day), medium (2 or 3 days) and long (4 or more days) length of stay

The duration of stay was not related to the reported presence of an OAU. The proportion of discharges on the same day as admission in 2006/2007 ranged from 17.3% to 60.2% on A&E based OAUs, 18.7% to 57.3% on ward based OAUs and 14.7% to 54.1% where no OAU was reported. Just over half of the episodes in both years were for children living in the most deprived quintile based on the Indices of Multiple Deprivation (52.1% in 2005/2006, 52.5% in 2006/2007).17 More than half of readmissions within 28 days occurred within 7 days (61.4% in 2005/2006, 57.8% in 2006/2007). The readmission rate within 7 days for all hospitals was 5.7% in 2005/2006 and 5.5% in 2006/2007. Eight hospitals recorded more discharges in 2006/2007 than in 2005/2006 and some increases were substantial (table 1 and supplementary online ‘Results’ file, table A). Most notable was an increase of 137% at hospital H which changed its recording policy to include more short term episodes in 2006/2007 than in 2005/2006. Its readmission rate also increased from 5.1% to 7.5%. There was one other large change in readmission rate, a reduction from 7.1% to 3.9% at hospital F, which was associated with only small reductions in the number of discharges and the rate of same day discharges. The next largest change in rate was a reduction of 1.4% at hospital L, and at eight hospitals the rate changed by 1% or less. Since much of the increase in discharges between 2005/2006 and 2006/2007 can be attributed to more comprehensive recording of ‘emergency day cases’ as admissions18 associated with the introduction of Payment by Results,13 further analysis of 2006/2007 only is presented.

Of the 1273 readmissions, 81% (n=1028) followed admission for 0–1 days, 12% (n=150) for 2–3 days, and 7% (n=95) for 4 or more days. Readmission rates were associated with same day discharges (table 3 and figure 1). The percentage of same day discharges from a hospital was highly correlated with its readmission rate (Kendall's taub=0.61, p=0.007). For breathing difficulty and feverish illness, the readmission rate also correlated with the proportion of same day discharges (breathing difficulty: Kendall's taub=0.83, p<0.001; feverish illness: Kendall's taub=0.50, p=0.023).

Figure 1

2006/07 Emergency readmission rates within 7 days and percentage of same day discharges. *Specialist children's hospitals.

Table 3

2006/07 Kendall's taub correlations between discharges and readmission rates per hospital

There was a strong association between the readmission rate and the total number of discharges in the year (Kendall's taub=0.71, p=0.002) (figure 2).

Figure 2

2006/07 Emergency readmission rates within 7 days and number of annual discharges. *Specialist children's hospitals.

For each of the conditions, the readmission rate for a hospital was also correlated with the total number of discharges in a year (breathing difficulty: Kendall's taub=0.72, p=0.001; feverish illness: Kendall's taub=0.61, p=0.006; diarrhoea: Kendall's taub=0.60, p=0.007) (table 3). The readmission rate was highest for diarrhoea in nine hospitals (and second highest in the remaining three hospitals), followed by breathing difficulty, with feverish illness having the lowest rate of readmission (table 4 and online ‘Results’ file, figure A).

Table 4

2006/07 Emergency readmission rates within 7 days disaggregated by diagnosis category

Most of the sample lived in the 40% most deprived areas in England. There was no significant association between the readmission rate and the percentage of admissions or readmissions living in the 10% most deprived areas (table 3). Ethnicity could not be examined as an independent variable because data were only available for 69.3% of the HES dataset, with wide variations between hospitals.

Discussion

Although HES data are cleaned prior to publication and duplicates removed, errors and missing data remain, some of which have been described in an audit of a relatively small sample.13 As the extent of inaccuracy is unknown, data should be treated with caution. However, a large dataset is reported here, which should minimise the impact of individual errors.

Readmissions have previously either been estimated across a wide range of conditions including surgical and other specialties, or studied in small samples that do not enable comparison between hospitals.9 This study is the first to our knowledge to have examined readmissions across a group of hospitals for common acute childhood conditions.

Clear associations between readmission rates and same day readmission and annual discharges were identified. These associations were also observed over time at hospital H which changed its coding procedure so that many more same day discharges were included in the year 2006/2007 than in 2005/2006. The number of discharges increased by 137%, from 1088 to 2583, and the number of readmissions increased by 245% from 56 (rate 5.1%) to 193 (rate 7.5%), providing additional evidence of the relationship between same day discharge and readmission.

Although subsequent emergency admissions are not necessarily related to the problem for which children were originally admitted (and may even be arranged), the relationship between discharge and readmission is consistent with clinical expectations. With earlier discharge there is more potential for a child's underlying condition or complications to develop and so require readmission.

It is possible that readmission rates are influenced by variations in patterns of health service use, including whether parents take their children directly to a hospital or first see a general practitioner. Although HES include data about method of admission, they were not sufficiently robust to enable analysis in this study.

Future studies could examine the relationship between the readmission rate and the number of discharges. Clinical staff in busier hospitals may be working under greater pressure, leading to more inappropriate early discharges. While hospitals admitting larger numbers of children are likely to have more staff and other resources, demand fluctuates seasonally and diurnally, and it may be problematic to match capacity to fluctuations in demand. In a US study, children with less complicated conditions experienced longer hospitalisations if admitted on days of high hospital occupancy, suggesting an association between workload pressures and quality of care.19 Future studies could explore relationships with seasonal disease and hospital occupancy.

Same day discharge and numbers of discharges are related to each other as well as to readmissions. The three hospitals that discharged more than 2500 children in 2006/2007 were among the five that discharged more than 50% of children on the day of admission; and the five lowest rates of same day discharge were at the five hospitals with the lowest annual discharges. This could suggest that hospitals develop services and policies for early discharge as a way of managing demand. It is also possible that the development of such services attracts additional demand.

The association between same day discharge and readmission raises questions about what is an appropriate rate of same day discharge and the trade-off between reducing unnecessary hospitalisation and increasing the chance of readmission. The five hospitals with readmission rates of 3.0% or less were also the hospitals that discharged the smallest proportion of children on the day of, or the day following, admission and the largest proportions of children on days 2 or 3 (table 2). This suggests that their low rates of readmission were at the expense of longer hospitalisations, with psychosocial implications for children and families and financial implications for the NHS. The group of five hospitals with the highest rates of readmission discharged 46% or more children on the day of admission. However, despite high rates of same day discharge at hospitals A (60.2%) and F (48.1%), their readmission rates were not proportionately high (5.4% and 3.9%, respectively) (table 1, figure 1 and online ‘Results’ file, table B). There is no obvious association with service design. The OAU in hospital A was attached to a dedicated paediatric A&E department suggesting that it might have more specialised staff and resources than other hospitals, but hospital F was a district general hospital with a ward based OAU. It is also notable that hospital F’s readmission rate reduced from 7.1% to 3.9% between 2005/2006 and 2006/2007 in association with only small reductions in the number of children admitted and the rate of same day discharges, indicating that other factors can also affect readmission rates. Because children were usually admitted to their local hospital, it was not possible to examine whether different community services affected readmission rates. While this study has demonstrated that readmission rates were associated with the number of annual admissions and the proportion of same day discharges, these data cannot identify the optimum rate of same day discharge or the risk to safety or quality indicated by readmissions. Further study of a larger number of hospitals and their associated community services, including children's community nurses, could explore how children can be safely discharged on the day of admission and so avoid unnecessary hospitalisation, while minimising the requirement for readmission.

Conclusions

Unnecessary overnight admissions to hospital are both an inefficient use of health resources and have potential for negative psychosocial consequences for children and their families. However, discharge decisions require the confidence of health professionals and parents that children's conditions will not deteriorate at home. Readmission rates are a measure of risk that should be taken into account when assessing the advantages and disadvantages of early discharge from children's emergency services. This study demonstrates that there are relationships between readmission and the overall number of children discharged, and the proportion discharged on the day of admission. These relationships could be explored further in future research to provide evidence to develop policy and practice that balance children's need for hospital care with avoidance of unnecessary hospitalisations.

References

Footnotes

  • Funding This independent study was commissioned and funded by the Department of Health Policy Research Programme which approved publication. The views expressed are not necessarily those of the Department of Health.

  • Competing interests None.

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