Discussion
During 2020, there were approximately 3700 fewer admissions, 36,000 fewer bed-days, and 100 fewer child deaths in PICU compared to the 2016–2019 average. There was a reduction in age–sex adjusted prevalence of unplanned PICU admission of 17 per 100,000 population in 2020 relative to the earlier period. The reduction in respiratory disease (in particular bronchiolitis) was the major contributor to the decrease in unplanned PICU admissions during 2020 with a striking absence of the expected winter peak. Conversely, there were significantly more admissions of patients with DKA compared to previous years. However, we did not observe an increase in morbidity or severity of illness using proxy measures such as probability of death using PIM3, length of PICU stay, advanced level of interventions and in-PICU mortality rates.
The virtual PICU systems (VPS) in the USA reported a 32% reduction in PICU admissions during April–June 2020 compared with 2017–2019 average admission figures for that quarter [
7]. However, data from July–December 2020 were not reported. We observed an identical 32% reduction in PICU admissions during the same time period, followed by 12% and 20% reductions in July–September and October–December, respectively. PICU bed-days for admissions between October and December 2020 in this report were 38% lower than the comparable pre-pandemic period because of the striking absence of a winter peak. The VPS report observed that admission numbers for the time period between January and March 2020 were comparable to that of the previous years. However, we report a 7% reduction during the same time period relative to 2016–2019. The first cases of COVID-19 were reported in January 2020 in the UK, and in February 2020 in the RoI. Lockdown rules were imposed in the UK and RoI in the last week of March 2020 [
15]. The ensuing reduction in bed activity that occurred in March 2020 is likely to have contributed to the difference. A report from Scotland that described data from their two PICUs, which are included in this dataset, also identified a similar reduction in PICU admissions with no substantial evidence of harm [
6].
Reduction in respiratory disease accounted for the majority of the difference in the number of PICU admissions and bed-days. Reports from North and South America have also revealed a similar trend of a significant reduction of respiratory viral illnesses [
7,
8]. The reasons are not entirely clear, although it may be related to the effectiveness of population-based non-pharmacological public health interventions. Several public health interventions were widely adopted in the UK and RoI during 2020 [
14,
15]. This included ‘stay-at-home orders’, social distancing measures, mobility restrictions, school closures, mask wearing and shielding of vulnerable populations [
14,
16]. Both the time periods with higher magnitude of reduction in PICU admissions and bed-days, i.e. April–June 2020 and October–December 2020, coincided with periods of national lockdown. Unlike April–June 2020, the stay-at-home orders were not accompanied by a national school closure during October–December 2020. However, there was a requirement for face covering, social distancing measures, better access to SARS-CoV-2 testing and isolation of symptomatic subjects. The direct impact of these public health interventions on COVID-19 has been reported [
17,
18]. The 2003 epidemic of severe acute respiratory virus (SARS) triggered an examination of the effectiveness of non-pharmacological interventions on other illnesses. There is some evidence to support the effectiveness of some of the strategies used during this pandemic in the spread of respiratory infections including influenza, upper respiratory tract infections, and respiratory syncytial virus infections, with a notable effect in reducing secondary attack rates in young children [
19,
20]. Given the reduction in the absolute number of child deaths in PICU reported in this study, as well as from other parts of the world [
21,
22], it remains to be seen whether some of the public health measures such as social distancing, wearing masks or face coverings and frequent hand-washing continue to be widely adopted and remain sustained in the community. It is as yet unclear how the low levels of exposure to respiratory pathogens among infants and children during the pandemic will impact on future seasonal and inter-seasonal respiratory viral surges. However, reports from Australia and elsewhere have raised this as a concern [
23].
An increase in DKA referrals was reported by one of the regional UK PICU transport services during the COVID-19 pandemic compared with previous years, despite a substantial reduction in overall referrals to the service [
24]. A study from the USA also reported a significant increase in patients with diabetes mellitus in PICU, but perhaps not necessarily with DKA [
7], while a report from across Canada identified a significant increase in both DKA and severe DKA presentations in new Type 1 diabetes mellitus [
25]. Our analysis confirms the trend of a recent increase in PICU admissions for DKA. However, it is unclear whether an association exists with SARS-CoV-2 infection, given that the increase in DKA admissions was first observed prior to the pandemic (October–December 2019), but remained sustained at higher levels during the pandemic. Delayed diagnosis of diabetes resulting in presentation as a DKA admission was raised as a concern [
26]. Comparisons of base deficit and length of stay of PICU admissions for DKA did not support an increase in severity of illness at PICU admission.
The VPS report described an increase in the proportion of trauma-related PICU admissions from 6.8% during April–June 2017–2019 to 9% of PICU admissions over the period April–June 2020, while the absolute numbers during this period in 2020 were lower than the previous years [
7]. Our observations were similar, but with a smaller non-significant increase from 2% of all PICU admissions in 2016–2019 to 2.3% in 2020. Again, there were fewer absolute numbers of patients admitted with trauma during 2020 and the differences are perhaps explained by a larger reduction in the non-trauma admissions, such as respiratory infections.
Altered health seeking behaviour resulting in delayed presentation has been implicated as a potential concern during the pandemic [
26]. Comparison of severity of illness on presentation to PICU during 2020 with data from the period 2016–2019, drawn from the analysis of the probability of death based on PIM3 calculations, PICU length of stay, need for advanced levels of critical care support, such as inhaled nitric oxide, renal replacement therapy or extracorporeal life support, did not reveal any evidence to support the hypothesis of delayed access to care. Analysis of data from several UK emergency departments confirms that delayed presentations were indeed rare [
27]. The results of this study support this conclusion, although timely and appropriate public health messages from relevant public health authorities may have helped.
Overall COVID-19 and PIMS-TS accounted for fewer than 500 PICU admissions in 2020. Just under half of all reported cases of PIMS-TS in the UK and ROI were admitted to PICU and we have established that these were predominantly for vasoactive medications rather than respiratory support [
2,
28]. By contrast, COVID-19 in adults resulted in excess of 25,000 critical care admissions in ICUs in England, Wales and Northern Ireland in 2020 [
29]. Some of the PICU capacity freed up by the reduced number of childhood admissions were then utilised to provide critical care for the adults [
5,
30]. Moreover, paediatric critical care staff were also redeployed in adult ICUs. This demonstrated an efficient and collaborative approach which was sensitive to the needs of the population, while preserving adequate paediatric critical care capacity.
Reduction in elective surgical activity was implemented in many parts of the world to free up capacity to cope with surge in COVID-19 related admissions. Several paediatric cardiac surgery programmes deferred most, if not all, elective activity [
31]. This report identified the impact on elective non-cardiac surgery activity, which was higher than the observed reduction in cardiac surgical activity. These numbers provide information about the anticipated need for additional PICU capacity when elective surgeries resume. In addition, differences observed between different surgical programmes and between different regions may provide important information for future pandemic planning.
We observed important differences compared to the US report from VPS. We observed no significant differences in probability of death using PIM3, or mortality during 2020, as opposed to a small but statistically significant increase in median PIM2 and mortality in the VPS dataset. The reasons for this are not entirely clear though this UK and RoI report analysed the whole of 2020 data and is population-based.
This study reports on a well-established registry covering the whole of the UK and the RoI. Data are validated and known to be of high quality [
9]. This dataset has established definitions and reports annually on numbers of admissions, severity indicators, specific diagnoses of interest and intervention related information as well as mortality. As a result we can be confident that the data we present reflect the front line experience of paediatric critical illness in the UK in 2020 and preceding years. We were able to present granular information presented in the form of monthly, as well as quarterly figures showing the evolution of the impact of the pandemic on PICU.
Nevertheless this report does have some limitations. Diagnostic category assignment is based on primary diagnosis as reported by clinicians in the participating units and some inconsistency within and between participating units may be present. Given that this data only represents designated PICUs, this data cannot be extrapolated to non-PICU hospitalisations or child deaths. A recently published study of national child mortality database in England reported that the all-cause mortality in children in 2020 (2264 deaths) was slightly lower that 2019 (2498 deaths) with a relative risk of 0.92 (0.87–0.98),
p = 0.009) [
32]. Nevertheless, comparative analysis of wider preceding time periods as well as from other countries is required before confirmation of this effect. This study does not include any other measures of health impact on children such as primary care attendances, hospitalisation, immunisation and other wider measures such as education, nutrition, fitness, poverty and mental health. These are also required to build an overall assessment to consider the risks and benefits posed by any large public health intervention in the future. An interrupted time series or regression discontinuity design analyses may be required to specifically analyse association between periods of the public health intervention measures and impact on PICU case-mix, outcomes during the pandemic. However, the timing and nature of interventions were variable at different countries. A regional approach was also pursued within some countries and these introduce significant barriers. As such our study merely describes the association between pandemic-related variables and the PICU case-mix, outcome differences, rather than test for causation.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.