Introduction

The impact of the COVID-19 pandemic on global health issues requires constant review of public health measures. In 2020, the United Kingdom implemented the first set of national restrictions for asymptomatic individuals after the recognition of COVID-19 transmissibility, mortality risk, and excess burden on the healthcare system. On 16 March, these included advice against non-essential travel, reducing contact between households, ban on visiting social venues, and ‘work from home’ initiative [1]. On 20 March, the hospitality sector was closed to dining-in and schools closed except to key workers’ children. Lastly, on 23 March, a ‘stay at home’ order was implemented and is recognised as the start of the first national lockdown, continuing until 12 May. From 13 May, significant restrictions remained in place with limited interaction permitted between households, despite a cautious return to work.

Internationally, varying levels of lockdown restriction has led to reductions in paediatric emergency department (ED) attendances of around 50% (24.9–88%) [2,3,4,5,6,7,8,9,10,11,12,13,14]. Many authors expressed concern about children who may be ‘lost’ to secondary services during lockdowns, particularly those with concerning or significant disease [4, 5, 10, 14,15,16]. Although children are deemed to be relatively unaffected by primary SARS-CoV-2 infection [17], pandemic measures/messaging are altering patterns of paediatric disease presentation. Recent studies during the 2020 winter months show significant declines in admissions for bronchiolitis, respiratory infections, and gastroenteritis compared to previous years [18, 19].

Few studies suggest an internationally comparable way to continuously report on patients not attending for hospital care. Our study sought to clarify the effects of the first UK lockdown on paediatric ED attendances and hospital admissions in Oxfordshire, UK, using ICD-10 (International Statistical Classification of Diseases and related health problems 10) codes from electronic patient records. We aim to identify diagnoses that were ‘missing’ during lockdown in 2020, compared to the previous 5 years (2015–2019).

Methods

Data was extracted from electronic healthcare records for all patients aged 0–15 years attending ED and any urgent admission under specialist services across Oxfordshire University Hospitals NHS Foundation Trust’s two sites (Horton General Hospital, Banbury, and John Radcliffe Hospital, Oxford). Data collection occurred 3 weeks prior to, inclusive of, and 2 weeks after the first national UK lockdown period (02/03/2020–26/05/2020), and matched periods in 2015–2019 (02/03/2015–26/05/2015, 07/03/2016–24/05/2016, 06/03/2017–23/05/2017, 05/03/2018–29/05/2018, and 04/03/2019–28/05/2019). Dates used for lockdown data were 23/03/2020–12/05/2020. The matched 49 days for 2015–2019 begin on 23/03/2015, 28/03/2016, 27/03/2017, 26/03/2018, and 25/03/2019. Matched dates included the same number of weekdays/weekends, beginning on the nearest Monday to the 2020 date.

Diagnoses using ICD-10 classification are assigned by clinical coders. ICD-10 codes were matched with chapter headings and specific diagnoses using https://icd.codes/icd10cm. To our knowledge, there was no substantial change to clinical coding during the 6-year study period. Due to software limitations, the first ten diagnoses for each patient were extracted. Patients are assigned a primary diagnosis, whereas concomitant diagnoses (e.g. viral illness, wheeze) and historical diagnoses add relevant context to the complexity of patients. Therefore, all diagnostic codes were included for most of our analyses. Two patients without a discharge date (from 2019) had length of stay censored at extraction; patients without discharge diagnoses are documented (0–9 patients/year; Supplemental Table 1).

A diagnosis of ‘COVID-19, virus identified’ (ICD-10 code U071) was made only five times during the 2020 lockdown period (never for dates outside of lockdown). COVID-19 was grouped under ICD-10 chapter 1 (infectious and parasitic diseases), and the sub-heading ‘other viral diseases’.

Additional data included date of birth, postcode, length of stay (LOS), and source of admission. Postcode was used for assessing Index of Multiple Deprivation Rank, calculated using the UK government online tool (http://imd-by-postcode.opendatacommunities.org/imd/2019).

Information used for clinical audit in May 2020 followed Trust procedures to access anonymised patient data. NHS Health Research Authority for ethics committee review and approval were not required.

Data were handled in Microsoft Excel (Office 365). Graphical presentation of data was prepared in GraphPad Prism (Version 9.0.0 for Windows, San Diego, California, USA). Statistical analyses were performed in GraphPad Prism and SPSS (Version 27.0. for Windows, IBM, NY, USA). Multiple testing corrections were performed where necessary. Non-parametric testing was used throughout; significance was taken as p < 0.05.

Results

ED attendances and inpatient admissions reduced during lockdown, without increased severity of admissions

We observed significant reductions in paediatric ED attendances and inpatient admissions across both Oxfordshire hospitals, compared with 2015–2019 (Fig. 1A–B). Pre-lockdown admissions reduced by 22.6% (p = 0.05) were lowest during the first week of lockdown (62.1% fewer, p < 0.0001) and maintained throughout (supplemental figure 1). These reductions were consistent across Oxfordshire (supplemental figure 2A-F) and were reduced in all age groups (age <1 year—48.4%, 1–5 years—67.2.4%, 6–10 years—53.3%, and 11–15 years—48.9%; supplemental figure 2G-J). The socioeconomic background of ED attendees was similar across all years 2015–2020, as assessed by postcode-derived index of multiple deprivation rank (supplemental figure 3).

Fig. 1
figure 1

Paediatric ED attendances and admissions during the first national lockdown period in 2020, compared to 2015–2019. The numbers of paediatric ED attendances (A), inpatient admissions (B), and paediatric inpatient admissions referred from ED (C) are shown by day for 2020 (purple) and compared to matched period (mean) 2015–2019 (grey). The proportion admitted from ED (with total ED attendances for the day as a denominator) is shown across the same period (D). For each day of admission, the resulting length of stay is calculated for each patient, and averaged across all patients admitted on the same day (E). Proportions of patients with more than 2 diagnoses on discharge from paediatric wards are shown, based on day of admission (F), with the histogram of the numbers of diagnoses received by year (G). National restrictions on asymptomatic individuals are shown (yellow) with the first national lockdown period (red). Lines of 2015–2019 represent simple linear regression with dotted line demonstrating 95% confidence. Lines for 2020 are non-linear segmental regression lines, using least squares fit and continuous hinge function

Full size image

We examined for surrogates of disease severity. Reductions in admissions from all sources were matched by reductions in admissions from ED alone (Fig. 1C); the proportion of ED attendees requiring admission was similar for each year (approximately 20%, Fig. 1D), and consistent between hospitals (supplemental figure 2K-L). Mean subsequent LOS, based on day of admission, was similar for patients admitted during lockdown (1.95 ± 0.338 days) compared with previous years (1.94 ± 0.120 days) (Fig. 1E). A surrogate of severity of illness was the number of diagnoses per patient: a mean of 71.6% received >2 ICD-10 diagnoses (mean 4.81 diagnoses/patient) in lockdown versus 44.5% (2.97 diagnoses/patient) in 2015–2019 (Fig. 1F–G). Greater proportions in 2020 were given ≥10 diagnoses (13.1% versus 1.3–4.1% in 2015–2919) (Fig. 1G; supplemental figure 4).

Diagnoses amongst inpatients were decreased in 8 of 20 ICD-10 chapters during lockdown

ICD-10 chapter headings provide diagnostic groups for children admitted in 2020 (Fig. 2; supplemental table 2). Significant reductions were observed for five of six of the most commonly diagnosed groups and eight of 20 chapter headings, compared with 2015–2019. Notable reductions in diagnoses were observed amongst respiratory diseases (62.4%, p < 0.0001), infectious diseases (58.9%, p < 0.0001) and injury/poisoning (52.7%, p < 0.0001). Only ‘neoplasms’ and ‘factors influencing health status and contact with health services’ increased in 2020 compared with previous years, by 75.9% (p = 0.0123) and 64% (p = 0.0004), respectively. When primary diagnosis is considered alone (supplemental figure 5), admissions within 11 of 20 chapters were significantly reduced in 2020. Neoplasms remained significantly increased throughout the lockdown period, whereas ‘factors influencing health status’ significantly decreased. Changes amongst these chapter headings were consistent week-to-week throughout the lockdown period (supplemental figure 6).

Fig. 2
figure 2

Changes in numbers of diagnoses made in 2020 compared to 2015–2019 by ICD-10 chapter. Diagnoses by ICD-10 chapter headings. Discharge diagnoses (maximum 10) are shown per week for the first national lockdown period (n = 7 weeks, 2020), compared to matched periods during the previous 5 years (n = 35 weeks, 2015–2019). Statistical results from Mann-Whitney U test. ns = not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Full size image

We explored relative changes in patient diagnostic groups admitted during lockdown (by chapter heading; Fig. 3). All diagnoses, and primary diagnoses, demonstrated reductions in the proportions of infectious and respiratory diseases. Similar proportions of non-specific diagnostic groups such as ‘factors influencing health status and contact with health services’ and ‘abnormal clinical or laboratory findings’ were observed amongst primary diagnoses in 2020 compared to previous years (Fig. 3, right panels).

Fig. 3
figure 3

Changes in relative proportions of diagnoses seen during the 2020 lockdown period. Differences in diagnoses by ICD-10 chapter headings for patients admitted during the first national lockdown period, compared to matched periods during the previous 5 years (2015–2019). Totals show mean numbers for the matched lockdown period (2015–2019) or total number for 2020. Shown for all diagnoses given on discharge (left panels) and ‘primary’ diagnosis only (right panels)

Full size image

Diagnoses ‘missing’ throughout the 2020 lockdown were predominantly infective illness or sequelae of infectious illness

To identify specific diagnoses that significantly differed in the seven weeks of lockdown, compared to the same period over the previous 5 years, we assessed differences in weekly admissions for each diagnosis (supplemental table 3). All diagnoses in 2015–2019 with a mean ≥1 or any diagnosis made in 2020 were included. There were 726.8 (20.4%) fewer diagnoses during lockdown versus 2015–2019 (n = 2853 in 2020 versus mean n = 3569.8 across 2015–2019). This included significantly more patients with neoplasms admitted in 2020 (Fig. 2), with notable increases for both benign and malignant neoplasms. Sub-group analyses showed a skew toward greater proportions of central nervous system (CNS) tumours, and secondary tumours diagnosed in 2020 (supplemental figure 7). Small but significant increases were observed in diagnoses of gram-negative sepsis, ‘unspecified’ appendicitis, and ‘other specified’ diabetes mellitus in 2020 (supplemental table 3). Sub-group classification was used to identify whether these may result from minor changes in diagnostic assignment for similar diseases (supplemental table 4). This showed no significant increase in any infectious disease subgroup, diseases of the appendix, or diabetes mellitus in 2020 compared to previous years. Amongst mental health diagnoses, significant reductions were observed for anorexia and the intentional self-harm subgroup, without any increased mental health diagnosis in 2020.

To identify the ‘missing’ diagnoses during lockdown, significant changes compared with 2015–2019 were categorised (Table 1). Net changes in numbers of diagnoses in 2020, compared with median numbers observed in 2015–2019, were used to demonstrate the degree of change. Diagnoses grouped as ‘likely related to pandemic screening’, ‘likely incidental finding or co-morbidity’, and ‘other diagnoses’ represented 24% of lockdown diagnoses and increased compared with 2015–2019 (Fig. 4A, right panel). ‘Infective illnesses’, ‘likely sequelae of infective illness’, ‘non-specific pains, aches and malaise’, and ‘accidental injury/poisoning’ had a net decrease in 2020. Net differences amongst ‘missing’ diagnoses totalled n = 761, more than accounting for the 726.8 fewer diagnoses in 2020 compared to the mean of 2015–2019 (Fig. 4A). Infective illnesses and ‘likely sequelae of infective illness’ accounted for 79.5% of diagnoses ‘missing’ from lockdown (Fig. 4B). Amongst infective illnesses, unspecified viral infections accounted for a third of missing diagnoses (Fig. 4C), whilst wheezing, nausea, and vomiting and unspecified asthma together constitute 72% of decreased diagnoses amongst sequelae of infective illnesses (Fig. 4D).

Table 1 ICD-10 diagnoses with significant changes during lockdown assigned to broad diagnostic groups
Full size table
Fig. 4
figure 4

Comparing diagnoses made in 2015–2019 to 2020, identifies ‘missing’ diagnoses of the first UK lockdown. The ‘missing’ patients are shown (white bar) as difference in mean annual diagnoses examined in 2015–2019 (A–left panel), against all diagnoses made in 2020 (A–middle panel). All ICD-10 diagnoses during the first national lockdown period in 2020 and matched dates in 2015–2019 are shown only where the diagnosis occurred annually (mean ≥ 1) in 2015–2019, or at least once in 2020, n = 1117 ICD-10 codes. Net changes for diagnostic groups with significant increases in diagnoses in 2020 and decreases (‘missing’) in 2020 are shown (A–right panel). The breakdown of ‘missing’ diagnoses (groups with a net significant decrease in diagnoses during the first 2020 lockdown period compared to 2015–2019) are shown by assigned group (B). The proportions of individual diagnoses which had a net decrease are shown amongst infective illnesses (C) and likely sequelae of infective illness (D)

Full size image

Discussion

Public health measures have varied widely internationally, even across the devolved UK nations. As a result, our data may not directly translate to every lockdown scenario. Our data included all acute paediatric admissions but precluded ED diagnoses, which are not classified by ICD-10 criteria. Therefore, extremely rare events or short ED attendances/observation periods will be under-reported in our data. Our study furthers others published to date (supplemental table 5), in performing a systematic examination of all paediatric inpatients and diagnoses using standardised ICD-10 criteria.

We reviewed 2843 diagnoses associated with 1416 admissions during the first lockdown period in 2020, compared with 12,458 admissions and 19,946 diagnoses across matched dates 2015–2019. We demonstrate decreased paediatric admissions and ED attendances throughout lockdown, with increased numbers of diagnoses per patient. Contrary to others [5, 12], the proportions of children admitted from ED were similar in all study years, suggesting comparable clinical severity of ED presentations across all 6 years. However, amongst those admitted a greater proportion received ≥10 diagnoses, suggesting an enrichment of children with complex/severe disease. Diagnoses which significantly increased during lockdown fit three main groups (1) increased documentation of co-morbidities/incidental findings, (2) pandemic screening, and (3) greater detection of ‘other’ diagnoses. Other diagnoses included benign and malignant neoplasms, similarly observed in German ED patients during their first national lockdown [5]. Urgent admissions increased for patients with malignancies, possibly due to reductions in face-to-face primary care capacity, and/or adaptations to tertiary oncology services during the pandemic, with elective admissions restricted.

Our findings have similarities to Germany’s first national lockdown, where paediatric diagnoses in Hanover were significantly reduced compared with 2019 for overall hospital admissions, communicable/non-communicable diseases, including intoxication [5]. In contrast, Dopfer et al. found greater proportions of ED patients were admitted, but with similar lengths of stay to 2019.

Significant reductions in infective diseases and personal injuries during international lockdown-like measures are recently reported [7,8,9,10,11,12,13]. Our data suggests 80% of ‘missing’ diagnoses during lockdown were infective disease or sequelae of infective illness. These most likely resulted from public health pandemic advice, with population adherence to infection control measures, compounded by restricting geographic movement and school closures throughout the first lockdown. Equivalent stringent measures in other countries also resulted in significant reductions in COVID-19 transmission as well as other circulating infectious diseases amongst children, including respiratory and gastrointestinal diseases [18, 19]. Similar to our findings, reductions in infectious diseases, their sequelae, and personal injuries were reported in paediatric presentations across the UK, Germany, Belgium, France, Italy, USA, South Africa, and Singapore [5, 7, 9, 12, 13, 18, 19]. We observed significant reductions in non-specific asthma and wheezing, despite historically high pollen counts during lockdown (Met Office UK, personal communication), reflected by increases in allergic rhinitis (p = 0.0828) and allergic asthma (p = 0.0613) amongst our data. Both asthma and wheezing are predominantly triggered by infection in children [20, 21], so were categorised as sequelae of infectious diseases. However, as these respiratory symptoms are also affected by air quality and pollution [22, 23], which significantly improved during lockdown [24, 25] alongside reduced environmental contacts due to a ‘stay at home’ policy, multiple factors may have contributed to the reduced asthma and wheeze admissions we observed.

With reduced exposure to infections, parent/carer confidence may increase in treating mildly unwell children with non-specific temperature/pain/aches/malaise for longer periods. Consistent with this hypothesis is an Italian study demonstrating reductions of 84–92% in paediatric ED patients triaged as having ‘minor injuries not requiring a doctor’ during lockdown [6]. Furthermore, a US study demonstrated increased high-severity ED attenders at triage during a state-wide stay-at-home period compared to the previous 3 years [12], suggesting low severity patients were kept at home. A small proportion of children may have not been brought to hospital due to documented anxieties surrounding hospital attendance during the pandemic [26], although current evidence suggests this group is small or negligible [27]. Reassuringly, a national study across Scotland found no difference in childhood inpatient mortality during lockdown, compared to the previous four years [14]. Therefore, reductions in infectious disease-related admissions likely result from (1) reduced infectious disease burden due to COVID-19 public health measures, (2) tolerating/monitoring children with mild and self-limiting illness at home due to greater parental supervision whilst locked down and/or information campaigns to avoid unnecessary healthcare utilisation, and (3) healthcare anxieties during the COVID-19 pandemic.

Non-specific pains/aches/malaise were significantly reduced during lockdown, many of which are likely benign and self-limiting. An Italian study also demonstrated significant reductions of children brought to ED with pain during lockdown [10]. We reiterate these authors’ concern that a proportion of these non-attenders may have concerning or red flag features, including safeguarding concerns.

Increased parental supervision, school closures, fewer opportunities for sports, and playground closures may account for the 9.6% reduction in accidental injuries/poisonings. Findings internationally support the reductions of all fractures [11, 28, 29], accidental injuries/poisoning [5, 8], and burns [30] during lockdowns. Our data corroborated a decrease in superficial injuries and lacerations, fractures, falls, and accidental poisoning, with a 74.5% reduction in childhood injuries. It remains of concern that some children who sustain injuries are not brought to medical attention but may later transpire to be more serious, requiring social/forensic or safeguarding investigations.

Recommendations

We propose communication pathways which

  1. (1)

    Establish the duration and nature of symptom morbidity, and level of parent/carer anxiety, prior to patients attending for emergency care, to provide feedback and reinforce health education;

  2. (2)

    Improve the immediacy of ‘telemed’ health advice, empowering parents/carers in managing mild and self-limiting illnesses at home;

  3. (3)

    Promote health information for parents/carers of children of all ages, including raising awareness of red flag symptoms, and alleviating healthcare anxieties throughout the pandemic.

Conclusions

Oxfordshire paediatric ED attendances and hospital admissions were markedly reduced by up to 62% during the first 2020 lockdown. By controlling for the activity within one region and using five historical years, our data encompasses the use of steady demographics and natural behaviours of the population when utilising emergency services over the 6-year study period, and better controls for year-to-year fluctuations.

By using hypothesis-independent standardised ICD-10 diagnostic codes, our findings demonstrate infectious diseases, or their sequelae, account for the majority (80%) of reductions in paediatric inpatients during lockdown. This is likely driven by reductions in infectious disease transmission due to public health measures. A further 20% of ‘missing’ diagnoses were accidental injuries/poisonings or non-specific pains/aches/malaise. We postulate that some ‘missing’ presentations may represent patients with concerning or evolving conditions, safeguarding concerns, and require attention but did not present to healthcare services. Hence, we emphasise that some children with concerning features of physical, social emotional, or mental illness may be missed.