A 5-year retrospective cohort study of unplanned readmissions in an Australian tertiary paediatric hospital

Billions of dollars of additional costs are incurred with unplanned hospital readmissions.^1,2 Rehospitalisation has both a physiological and psychological effect on patients and their family or carers.^3–5 In Australia, the unplanned hospital readmission rate is defined as the percentage of unplanned readmissions to the same hospital within 28 days of discharge.⁶ However, most studies in the literature use 30 days to measure readmission rates.^7–12 All-cause unplanned hospital readmission rates in children range from 3.4% to 22.4% based on 30-day to 2-year follow-up periods.^1,2,8–14 An American study¹⁵ examined 30-day unplanned hospital readmissions across 72 children’s hospitals over a 12-month period, finding that the readmission rate was 6.5% and that the three most common principal diagnoses associated with readmissions were seizure, bronchiolitis and pneumonia. Furthermore, the most common reason for readmission in nine of 10 readmissions was the same diagnosis as the index admission.¹⁵ For condition-specific unplanned readmissions, prevalence ranges from 19% to 31% for mental health conditions based on 12-month measurements,^16–19 from 4.5% to 38% for respiratory diseases based on 28-day to 1-year measurements^20–25 and from 0.3% to 27.8% for general surgeries based on 7-day to 1-year measurements.^26–34

There is limited published literature measuring all-cause paediatric unplanned hospital readmissions in Australia. Five Australian studies were identified with a focus on specific health condition-associated readmissions; these studies examined readmissions across mental health conditions,¹⁶ asthma,^24,25 term live-born infants³⁵ and paediatric intensive care patients.³⁶ Unplanned readmission rates ranged from 0.8%³⁵ to 38%.³⁶ In Western Australia (WA), the prevalence of 28-day all-cause readmissions to the same hospital across all metropolitan public health services almost doubled from 2.1% in 2010–11 to 3.9% in 2014–15.^37–41

The aim of the present study was to characterise the frequency and nature of all-cause unplanned hospital readmissions in a tertiary paediatric hospital in WA from 2010 to 2014. Specific objectives were to: (1) determine the 5-year prevalence in paediatric admissions and characteristics of unplanned hospital readmissions; (2) examine the time interval between the paediatric unplanned hospital readmission and the initial hospitalisation; and (3) characterise principal index admission diagnoses associated with frequent paediatric readmissions.

A retrospective cohort descriptive study was conducted at the Princess Margaret Hospital for Children (PMH, which was relocated and renamed to Perth Children’s Hospital on 10 June 2018), a 220-bed public acute care hospital with approximately 250 000 patient visits (in-patient and out-patient) each year. At the time of the study, the PMH was the only tertiary paediatric facility providing care for children and adolescents within WA.⁴²

Patients of all ages discharged from the PMH between 1 January 2010 and 31 December 2014 were extracted as an electronic administrative in-patient dataset from the WA Hospital Morbidity Data Collection (HMDC). Emergency department presentations and/or emergency department short-stay unit admissions were excluded from the study. In addition, deceased patients and discharges of mothers attached to neonatal patients who were transferred to the PMH from maternity hospitals were excluded from the study, as were patients whose parent, guardian or carer discharged them against medical advice and those patients transferred to other hospitals (because of incomplete hospitalisation and variations in the discharge process).

Ethics approvals were sought from the human ethics research committees of the PMH, Curtin University and Department of Health, WA.

Hospital readmission is measured from an index admission, the first hospitalisation for a specific clinical condition, to the subsequent unplanned admission. Subsequent admissions that are related to the index admission and occurred unexpectedly within a specified time interval are considered unplanned hospital readmissions. The identification of unplanned hospital readmissions in this study was based on the combination of admission type (emergency) and the principal diagnosis of the subsequent admission following the index admission.

Variables extracted from the HMDC included demographic and clinical information for each patients. Demographic variables included age, sex and residential postcode. Clinical data included the date of admission, date of discharge, principal discharge diagnosis, date of readmission, date of discharge from the readmission and the principal diagnosis of readmission. Age was coded into six groups: infants aged <12 months, preschool children aged 1–4 years, primary school children aged 5–8 years, late primary school children aged 9–12 years, lower secondary school children aged 13–15 years, upper secondary school children and young adults aged ≥16 years. It is also worth noting that the age limit to be admitted to PMH is <16 years unless special permission is granted by the hospital executives for patients with a pre-existing medical condition.⁴²

Residential postcodes of each patient were converted into Socio-economic Indexes for Areas (SEIFA), a broad definition of relative socioeconomic status regarding people’s access to material and social resources and their ability to participate in society.⁴³ The Index of Relative Socio-economic Advantage and Disadvantage (IRSAD) was used in this study with scores that ranged from a percentile 0 to 100%. Lower IRSAD scores relate to households with a greater socioeconomic disadvantage.⁴³

In the present study, 4403 condition-specific principal diagnoses based on the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM)⁴⁴ diagnoses codes were initially extracted from the HMDC. These ICD-10-AM codes were then grouped according to their first three characters, resulting in 494 groups.⁴⁴

The date of discharge from the index admission (Day 0) until 30 days after being discharged (Day 30) was examined in this study.

Identification of the top principal index admission diagnoses was based on readmission counts.

Data were analysed using SPSS version 23.0 (IBM Corp., Armonk, NY, USA). The baseline prevalence of index admissions and readmissions was calculated on discharge-based and patient-based data because some patients had more than one index admission and readmission during the 5-year period. If the patient had more than one unplanned readmission after being discharged within the 30 days, only the first readmission was included for the discharge-based analysis. Patient characteristics are described as the mean ± s.d. or median with interquartile range (IQR) for continuous variables, and as counts and percentages for categorical variables.

In all, 137 621 discharges from the PMH between 2010 and 2014 involving 75 524 individual patients were extracted from the HMDC. Seven hundred and fifteen discharges, associated with maternity issues and attached to neonatal patients were excluded from the study, as were 192 deceased patients and 136 patients who left the hospital against medical advice. A further 2264 discharges were transfers to other hospitals and were therefore also excluded. Thus, 134 314 discharges of 73 132 patients were included in the final analysis. Of the 73 132 patients, 740 experienced more than one 30-day unplanned hospital readmission; of these 740 patients, 403 had two readmissions, 151 patients had three readmissions and 186 patients had four or more unplanned readmissions. Of the 134 314 discharges, 1479 were admitted with a principal diagnosis of T81 (complications of procedures, not elsewhere classified). In particular, patients were admitted to the PMH when they experienced postoperative complications following an initial surgical procedure that had been performed at other public or private hospitals. The top two postoperative complications were haemorrhage and haematoma (n = 1024) and wound infection (n = 258). The remaining 197 complications were varied.

Based on analysis of all discharges, the number of hospitalisations was similar across the 5-year data collection period (Table 1). The total number of discharges exceeded 24 900 each year, with a mean age of 6.2 ± 5.1 years. Regarding patient-based analyses (Table 1), the number of patients hospitalised in each year of the study was similar, with a mean age of 5.7 ± 5.0 years. Overall, one-third of patients were aged from 1 to 4 years, and 1.5% of patients were aged >16 years. There were 16% more male than female patients.

Table 1.  Summary of all discharges between 2010 and 2014 from the Princess Margaret Hospital for Children
All percentages are based on the total for a year. Data were analysed on the basis of both discharges (where patients may have multiple visit) and patients. Unless indicated otherwise, data are presented as the mean ± s.d. or as n (%)

Based on analysis of all discharges, readmission rates ranged from 2.97% to 3.03%, with more male than female patients featuring in both the with- and without-readmission groups (Table 2). From a patient-based analysis perspective (Table 2), 4.55% of patients (n = 3330) experienced 30-day unplanned hospital readmissions, which ranged from 3.86% to 5.04% over the 5-year data extraction period. The mean age of the with-readmission group was 6.3 ± 5.4 years, compared with 5.7 ± 4.9 years in the without-readmission group. Length of stay of the index admission was almost double in the with- versus without-readmission group (4.7 ± 14.1 vs 2.4 ± 6.6 days respectively). The mean SEIFA percentile was similar in these two patient groups.

Table 2.  Summary of patient characteristics for those with and without unplanned hospital readmissions within 30 days of discharge
Data were analysed on the basis of both discharges (where patients may have multiple visit) and patients. Unless indicated otherwise, data are presented as n (%). IQR, interquartile range; LOS, length of stay of the index admission; SEIFA, Socio-economic Indexes for Areas

Fig. 1 shows the time interval from the discharge date of the index admission to the date of readmission of all-cause unplanned hospital readmissions. Each day from Day 0 to Day 7 accounted for more than 5% of unplanned hospital readmissions (Fig. 1a). The highest number of readmissions (532; 16%) occurred on Day 1, followed by Day 2 (n = 345; 10.4%) and Day 3 (n = 251; 7.5%). Of note, 208 readmissions (6.2%) occurred on Day 0. There were minimal differences in the rate of readmissions on Days 28, 29 and 30 (0.5%, 0.7% and 0.5% respectively). Fig. 1b shows the cumulative percentage of unplanned hospital readmissions from Day 0 to Day 30. More than half the readmissions (52%; n = 1732) occurred between Day 0 and Day 5, with 73.6% (n = 2450) occurring in the first 11 days (Days 0–10) and 86.1% (n = 2868) occurring in the first 16 days (Days 0–15).

Fig. 1.  Time from discharge after the index admission to the subsequent unplanned readmission showing (a) the percentage and (b) cumulative percentage of unplanned readmissions on each day after discharge.

The distribution of time intervals from discharge to unplanned readmission varied between each index admission diagnosis (Fig. 2). Nearly all the readmissions (98%) for the diagnosis code J05 (croup and epiglottitis) occurred by Day 5 after discharge, whereas readmissions for the G47 (sleep disorders) and J21 (acute bronchiolitis) diagnoses primarily (≥95%) occurred by Day 15. Readmissions related to other diagnoses (e.g. K35 (acute appendicitis), R10 (abdominal and pelvic pain) and R56 (convulsions) were spread across the 30 days after discharge (Fig. 2). Fig. 2a shows the percentage of condition-specific associated readmissions from Day 0 to Day 30, whereas Fig. 2b shows the cumulative percentage of readmissions.

Fig. 2.  Time from discharge after the index admission to the subsequent unplanned readmission for three principal diagnoses at index admission showing (a) the percentage and (b) cumulative percentage of unplanned readmissions on each day after discharge. G47, sleep disorders – obstructive sleep apnoea syndrome requiring tonsillectomy and/or adenoidectomy; K35, acute appendicitis leading to appendectomy; J05, croup and epiglottitis.

The top 10 principal index diagnoses associated with unplanned readmissions are given in Table 3. The most frequent diagnosis was G47 (sleep disorders – obstructive sleep apnoea (OSA) syndrome requiring tonsillectomy and/or adenoidectomy), followed by K35 (acute appendicitis leading to appendectomy) and J21 (acute bronchiolitis). The three most common reasons for readmission are summarised in Table 3. The most common readmission diagnosis was the same diagnosis for the top 10 index admission diagnoses, especially J45 (asthma; 89%). Readmissions related to postoperative complications of initial surgical procedures ranged from 40.5% to 85.7%, with diagnoses including G47 (OSA), J35 (chronic diseases of tonsils and adenoids) and K35 (acute appendicitis).

Table 3.  Three most common reasons for 30-day unplanned readmission for the top 10 principal index admission diagnoses

This study provides a baseline of the prevalence and characteristics of all-cause unplanned readmissions to the PMH using a large administrative dataset. The study identified a rate of 3.03% readmissions based on discharge analysis and a rate of 4.55% based on patient analysis. Studies conducted in other countries examining all-cause 30-day paediatric readmissions found considerable variations in prevalence rates. In terms of discharge-based analysis, three studies reported readmission rates of 10.3%,¹⁰ 8.8%¹³ and 6.5%.¹⁵ In another three studies, readmission rates based on patient analysis were found to be 3.4%,¹² 4.5%⁹ and 18.7%.¹¹ Two studies described readmission rates of 6.5%⁸ and 8.8%⁴⁵ based on both discharge and patient analyses. Of these eight studies, six examined readmissions to both the same and different hospitals.^{8,9,12,13,15,45} The two studies^10,11 that the examined same-hospital readmissions reported much higher rates than found in this study. As highlighted in a recent systematic review,⁴⁶ caution is required when comparing readmission rates in the literature, because this is dependent on whether readmissions are based on discharges or patients, and whether readmissions are to the same or different hospitals.

Identification of the frequent principal index admission diagnoses associated with unplanned readmissions is commonly based on total counts.¹⁵ The top 10 diagnoses in this study (Table 3) are consistent with those found in an American study,¹⁵ except for one diagnosis, seizure. An Australian government report stated more than 30% of patients experienced an unplanned readmission following their initial surgical procedure.⁷ The most frequent surgical procedure was tonsillectomy and/or adenoidectomy, which is consistent with our findings in a paediatric population.⁷ The present study also supported the findings of Berry et al.¹⁵ that most readmissions are for the same diagnosis or complications associated with the initial admission. The top diagnoses identified in the present study could be targeted to improve continuity of care at discharge and therefore to reduce the readmission rate.^3–5

Most studies in the literature use a period of 30 days to measure unplanned hospital readmissions, although some studies have used 7 days^30,45,47 to measure readmissions. Berry et al.¹⁵ found that 39% of all-cause readmissions occurred in the first 7 days after discharge and that 61.6% occurred in the first 14 days. In the present study, 52% of readmissions occurred within the first 5 days after discharge, 73.6% occurred in the first 10 days and 86.1% occurred in the first 15 days. The difference in unplanned hospital readmission rates between Days 28 and 30 was 0.2%, confirming that the findings across both periods are comparable. Fig. 2 clearly shows that some diagnoses are captured as Day 30 readmissions (K35, R10 and R56), whereas others are more accurately characterised as Day 15 readmissions (G47 and J21) or Day 5 readmissions (J05). These findings support the Australian Independent Hospital Pricing Authority’s⁴⁸ emphasis on the need to establish appropriate time intervals to measure readmission according to specific health conditions.⁴⁹ This will lead to accurate determination of the prevalence and true cost of readmissions.^50,51 This is particularly important because of the need to improve efficiencies in resource utilisation within the healthcare system, as directed by funders, including government and private health insurance.^48,52

The present study is limited by the fact WA has one tertiary paediatric hospital (the PMH). The present study did not include index admissions to the PMH but then readmitted to non-paediatric hospitals or admission and readmissions to non-paediatric hospitals with paediatric wards. A future study incorporating WA linkage data would address these omissions and enable patients admitted to both non-paediatric hospitals and then the PMH with unplanned hospital readmissions to the same or a different hospital to be captured. This study was a 5-year audit of an administrative database; individual in-patient files were not accessed to capture specific clinical information.

The present study is the first to date to examine paediatric all-cause unplanned same-hospital readmissions in WA based on an in-patient administrative dataset of all PMH discharges from 2010 to 2014. The study found a higher same-hospital readmission rate compared with the WA metropolitan public hospitals,^37–41 but lower than the rates in the US^8–11, UK¹³ and Canada.¹² The present study identified that the commonly used 30-day readmission follow-up period requires review because there are differences in the time intervals from discharge to readmission among diagnoses. It is critical to establish the most suitable measurement for readmissions for the accurate determination of the prevalence and true costs of readmissions. Investigations and strategies to reduce the occurrence of 5-day readmissions because of their high prevalence and readmissions on the discharge date are urgently needed. Identification of top index admission diagnoses for paediatric readmissions is essential to improve continuity of care at discharge with the aim of reducing unplanned hospital readmissions.

None of the authors has any conflicts of interest to disclose.