Introduction

By the height of the SARS-CoV-2 (COVID-19) pandemic, it had become apparent that children would be spared the brunt of the mortality and physical morbidities associated with the worldwide infection [1, 2]. Most children infected with SARS-CoV-2 were asymptomatic or exhibited mild disease [3]. Less than 10% developed severe disease that included myocardial dysfunction, shock, paediatric acute respiratory distress syndrome (PARDS), altered mental status, and Multisystem Inflammatory Syndrome in Children (MIS-C). Initial reports in paediatrics called attention to MIS-C with few reporting on the trajectory of isolated SARS-CoV-2 related PARDS [4].

Systematic investigation of the pathophysiologic effects of SARS-CoV-2 related lung disease in children has been limited with no comprehensive multicentre studies published. In a single-center study, Perk and colleagues showed that children who tested positive for respiratory viral pathogens other than SARS-CoV-2 were more likely to be hypoxemic than those with COVID-19 infection [5]. In contrast, Chao et al. described a higher than previously appreciated rate of severe disease in children admitted to the pediatric intensive care unit (PICU) with SARS-CoV-2 infection [6]. Obesity and congenital heart disease have been associated with more severe disease [7, 8].

The infrequent occurrence of SARS-CoV-2 acute respiratory failure in children has limited our capacity to comprehensively study the disease and compare it to known problems associated with PARDS [9, 10]. Pulmonary supportive strategies center on symptomatic management of PARDS according to the Pediatric Acute Lung Injury Consensus Conference (PALICC) guidelines [11, 12]. Paediatric SARS-CoV-2 specific management evolved over the pandemic and was based on knowledge acquired in adult patients with acute respiratory distress syndrome (ARDS) [13].

To date, no study has comprehensively described children with isolated SARS-CoV-2 acute respiratory failure who required mechanical ventilation. Understanding this disease and contrasting it to what is known about PARDS is an important step in knowledge development in the field. Here, we characterize a large cohort of children with SARS-CoV-2 acute respiratory failure, without MIS-C, and compare them to a pre COVID-19 pandemic cohort of children requiring invasive mechanical ventilation for acute respiratory failure.

Methods

SARS-CoV-2 patients were prospectively enrolled in a COVID-19 observational study that was associated with the Prone and Oscillation Paediatric Clinical Trial (PROSpect), an international randomized controlled trial investigating supine versus prone positioning and conventional ventilation versus high-frequency oscillatory ventilation (HFOV) in paediatric patients with high-moderate to severe PARDS (ClinicalTrials.gov: NCT03896763). During the pandemic, PROSpect sites were well-positioned to enroll SARS-CoV-2 positive patients into an observation-only study as the research infrastructure support was in place and many sites were required to pause their non-COVID-19 related research. Thirty-three PROSpect sites were eligible to participate in the observational study, as they were institutional review board- and system-approved and open for COVID-19 patients in their unit.

We enrolled paediatric patients ≥ 2 weeks of age (≥ 42 weeks post gestational age) and < 18 years of age who were mechanically ventilated with chest imaging consistent with acute pulmonary parenchymal disease and a confirmed positive SARS-CoV-2 polymerase chain reaction or antibody test, who did not exhibit MIS-C associated with COVID-19 [4, 14] and were not enrolled into PROSpect.

The matched cohort of non-COVID-19 patients was derived from the RESTORE (Randomized Evaluation of Sedation Titration for Respiratory Failure) study of protocolized sedation versus usual care in paediatric patients mechanically ventilated for acute respiratory failure (ClinicalTrials.gov: NCT00814099) [15]. While RESTORE enrolled subjects in 2009 through 2013, it is well documented that clinical management of PARDS has not changed during this time [12, 16, 17].

The COVID-19 observational study was approved by the University of Pennsylvania institutional review board (IRB) for all U.S. sites (‘Prone Oscillation Pediatric Clinical Trial’; IRB #831295; annual approval, most recent August 8, 2023) and at the local institutional level for international sites. Data collected during the RESTORE study was with IRB approval at each participating site. Written informed consent from the legal guardian of each enrolled patient was obtained for both cohorts. All aspects of the PROSpect and RESTORE trials were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional or regional) and with the Helsinki Declaration of 1975.

Functional status in the COVID-19 cohort was assessed using the Paediatric Cerebral Performance Category (PCPC), Paediatric Overall Performance Category (POPC), and Functional Status Scale (FSS) [18, 19]. Quality of life was assessed using the chronological age-appropriate Paediatric Quality of Life Inventory (PedsQL™; Version 4.0 Generic Core Scales for subjects ≥ 2 years; Infant Scales for subjects < 2 years) [20, 21] as reported by parent/guardian at 12 months post-PICU discharge. Functional status in the RESTORE cohort was assessed using the PCPC and POPC.

Statistical analysis

Each COVID-19 patient was matched with up to four unique, randomly selected patients from the RESTORE study based on age category (< 4.99, 5–11.99, 12–15.99, and 16–17.99 years of age), PARDS category based on worst oxygenation index (OI) or oxygen saturation index (OSI) on days 0–1 of mechanical ventilation (severe (OI ≥ 16.0 or OSI ≥ 12.3), moderate (OI of 8.0–15.9 or OSI of 7.5–12.2), mild (OI of 4.0–7.9 or OSI of 5.0–7.4), or at risk (OI < 4.0 or OSI < 5.0)) [12], and presence of organ dysfunctions (respiratory, cardiovascular, hematologic, neurologic, and renal) on days 0–1. Matching was performed using these variables to control for severity of illness. Categorical age groups were chosen based on differences in COVID-19 vaccine eligibility. As all matching variables were categorical, exact matching was performed.

The primary outcomes were ventilator-free days (VFD) through day 28 and in-hospital mortality at day 28. Patients who died by day 28 were assigned zero VFD. Comparisons of baseline characteristics, PICU course, and outcomes between the two cohorts (COVID-19 and RESTORE) were performed using stratified Wilcoxon rank-sum tests for continuous and ordinal variables and stratified exact tests for binary and nominal variables, adjusting for the matching factors. For the primary outcomes, we report the unstratified Hodges-Lehmann location shift for VFD and the stratified odds ratio for 28-day mortality with 95% confidence intervals. Patients with the same combinations of age, PARDS, and organ dysfunction categories were combined into the same strata for more efficient analyses. As we considered this matched cohort study descriptive in nature, we did not adjust for multiple comparisons. We explored whether obesity in mechanically ventilated paediatric patients with SARS-CoV-2 infection is associated with worse outcomes by comparing the primary outcomes between obese and non-obese patients within the COVID-19 and RESTORE cohorts separately. Weight categories were determined using weight-for-height z-scores based on World Health Organization growth charts for children aged < 2 years and body mass index-for-age z-scores based on Centers for Disease Control and Prevention growth charts for children aged 2 years and older. Statistical analyses were performed with SAS software, Version 9.4 (SAS Institute, Cary, NC) using two-sided 0.05 level tests.

Results

From May 29, 2020 to June 6, 2022, 325 COVID-19 positive patients were identified at 31 sites participating in the COVID-19 observational study. Of the 272 COVID-19 positive patients not enrolled, 98 were not intubated because of COVID-19 related acute respiratory failure, 37 had parent/guardian decline consent, 22 were not in the study age range, 21 were enrolled in the PROSpect trial, 18 had parent/guardian unavailable, 3 exhibited MIS-C, and 73 were not enrolled for other reasons (e.g., language barrier, system issue). Fifty-three patients were enrolled in the COVID-19 observational study across 21 sites in the United States, Canada, and Australia. These patients were matched with 195 patients from the RESTORE trial, including 102 in the RESTORE sedation protocol group and 93 in the usual care group. Of the COVID-19 patients, 42 matched exactly with 4 RESTORE patients, 6 matched with 3 RESTORE patients, 4 matched with 2 RESTORE patients, and 1 matched with only 1 RESTORE patient.

Cohort characteristics

Baseline demographic and medical history information for the PROSpect COVID-19 observational cohort and the RESTORE cohort are shown in Table 1. Both cohorts had predominantly direct PARDS with pneumonia as the leading diagnosis. Significantly more patients in the COVID-19 cohort were obese as compared to the RESTORE cohort (43% vs 19%; P = 0.001). More COVID-19 patients were Hispanic/Latino (37% vs 21%; P = 0.048). Significantly more COVID-19 patients were cognitively or functionally impaired at baseline as assessed by PCPC > 1 (57% vs 33%; P = 0.004) or POPC > 1 (64% vs 36%; P < 0.001). The paediatric risk of mortality was lower in the COVID-19 cohort (median 1.2% vs 4.9%; P < 0.001). Additionally, those with COVID-19 induced PARDS were more likely to have an underlying respiratory condition (e.g., asthma, obstructive sleep apnoea) or neurologic condition (e.g., seizure, neuromuscular disorder, cerebral palsy) or Trisomy 21.

Table 1 Baseline Characteristics of PROSpect COVID-19 Observational Cohort and RESTORE Matched Cohort According to Group
Full size table

PICU course

The PICU course for the COVID-19 and RESTORE cohorts is displayed in Table 2. COVID-19 patients had more severe PARDS during hospitalization compared to RESTORE patients (severe PARDS 75% vs 67%; P = 0.01). Only 4% of the COVID-19 cohort were managed with HFOV as compared with 17% of the RESTORE cohort (P = 0.01). Those with COVID-19 related PARDS were more likely to receive neuromuscular blockade for the entire duration of days 1 and 2 (36% vs 17%; P = 0.002), receive vasoactive medications (62% vs 48%; P = 0.008), and be cannulated for extracorporeal membrane oxygenation (ECMO; 15% vs 4%; P = 0.006). Sedative use was higher in the COVID-19 cohort with an increased mean daily opioid dose (median; 2.1 vs 1.3 mg/kg; P < 0.001), cumulative opioid dose (26.3 vs 14.9 mg/kg; P = 0.02), and number of different sedative classes received (4 vs 3; P < 0.001) (Table 2). In the COVID-19 cohort, administration of dexmedetomidine, propofol, ketamine, and barbiturates were increased, but benzodiazepine use was lower compared to the RESTORE cohort.

Table 2 PICU Course of PROSpect COVID-19 Observational Cohort and RESTORE Matched Cohort According to Group
Full size table

Management strategies

Specific management of the COVID-19 cohort is summarized in Table 2. Common interventions included the use of anti-infective agents (81%), anticoagulants (68%), and steroids (53%). One COVID-19 patient received one vaccine dose prior to admission. On enrolment, static compliance was measured during an inspiratory hold in 29 patients: median 15.0 mL/cm H2O (interquartile (IQR) 8.1–23.4) or 0.37 mL/cm H2O/kg (0.31–0.43). Of these, 28 patients (97%) had direct lung injury.

Respiratory support and outcomes

Ventilator-free days through day 28 were median 17.6 (IQR 3.3–22.3) in the COVID-19 cohort and 20.3 (14.7–23.7) in the RESTORE cohort (unstratified Hodges-Lehmann location shift –2.4 days [95% CI, –‍4.9 to –0.5]; stratified Wilcoxon P = 0.003; Table 3). Thirteen patients (25%) in the COVID-19 cohort were assigned zero VFD compared to 21 (11%) in the RESTORE cohort (P = 0.02). The duration of assisted breathing (through 90 days) was longer in the COVID-19 cohort with a median of 14.0 days (7.8–52.8) vs 9.5 days (5.6–16.9) in the RESTORE cohort (P = 0.004). A higher proportion of COVID-19 PARDS patients received non-invasive respiratory support pre-intubation and post-extubation (66% and 74% vs 45% and 43%; P = 0.003 and P < 0.001, respectively).

Table 3 Outcomes of PROSpect COVID-19 Observational Cohort and RESTORE Matched Cohort According to Group
Full size table

No in-hospital mortality differences were seen at 28 days (8% in the COVID-19 cohort vs 6% in the RESTORE cohort; stratified odds ratio 1.14 [exact 95% CI, 0.24–4.20]; P > 0.99; Table 3) or 90 days. Among survivors, COVID-19 patients were ventilated longer through day 28 (median 9.5 vs 7.3 days; P = 0.003) and had longer PICU (12.6 vs 11.2 days; P = 0.02) and hospital (19 vs 17 days; P = 0.02) lengths of stay. No difference was seen in change in cognitive or functional outcome between cohorts.

Obesity status

Compared to 30 non-obese COVID-19 patients, 23 obese patients with SARS-CoV-2 infection did not experience significantly fewer VFD (median 18.0 (IQR 0–22.2) vs 17.2 (9.8–22.4) days; P = 0.50) or higher in-hospital mortality at 28 days (3 (13%) vs 1 (3%); P = 0.31). Similarly, 37 obese patients without SARS-CoV-2 infection did not experience significantly fewer VFD (median 19.3 (IQR 14.0–23.0) vs 20.5 (16.0–23.8); P = 0.68) or higher 28-day hospital mortality (2 (5%) vs 10 (6%); P > 0.99) compared to 158 non-obese patients in the RESTORE cohort. In both cohorts, there was no difference in race and ethnicity by obesity status.

Post-hospitalization follow-up

Among survivors in the COVID-19 cohort (n = 47), 17 parents/guardians were not contacted for follow-up, including 12 for whom contact information could not be obtained, 3 outside the United States, 1 with guardianship issues, and 1 who refused follow-up. Nineteen parents/guardians provided follow-up data at 12-months post-PICU discharge, while 11 did not respond to multiple contact attempts.

Of the 19 patients with follow-up data, 11 (58%) had cognitive impairment (PCPC > 1) and 12 (63%) had functional impairment (POPC > 1) at baseline per medical record review, 14 (74%) had cognitive and functional impairment at follow-up per parent-report, and 10 (53%) had a worsening of cognitive and functional outcome at follow-up from baseline. At follow-up, median FSS score was 12 (IQR 8–15) for 17 patients, suggesting mild to moderate dysfunction per domain. Mean PedsQL total score at follow-up was 69.1 (standard deviation (SD) 22.6), mean physical health summary score was 75.3 (SD 26.6), and mean psychosocial health summary score was 66.1 (SD 23.8). Total scores in healthy children average 82.7 (SD 15.4), physical health summary scores average 84.5 (SD 19.5), and psychosocial health summary scores average 81.7 (SD 15.2) [22]. PedsQL total and physical health summary scores could not be calculated for 5 COVID-19 patients (4 with PCPC of 4 or 5 (severe disability or coma/vegetative state)) and psychosocial health summary score could not be calculated for 2 patients (1 with PCPC of 5) due to parents responding “does not apply to my child” or not answering more than half of the relevant questions.

Discussion

While much has been learned about COVID-19 illness since early 2020, knowledge of severe COVID-19 infection in children remains limited. Our study imparts new information regarding the effects of severe SARS-CoV-2 infection in a largely unvaccinated paediatric population as compared to a matched cohort of mechanically ventilated patients with acute respiratory failure. Paediatric patients with COVID-19 related acute respiratory failure were more likely to have a comorbid respiratory or neurologic condition, be obese, be Hispanic/Latino, and have cognitive and/or functional impairment on PICU admission. Compared to children with non-COVID-19 related acute respiratory failure, their course of illness was also more severe. They experienced longer durations of mechanical ventilation and length of PICU stay.

It is unclear whether the course of critical illness was primarily associated with direct effects of COVID-19 infection or whether management strategies, including the need for strict airborne isolation, especially pre-healthcare worker vaccination, including limited entry into patient rooms contributed. In the current study, patients with COVID-19 were more likely to receive neuromuscular blockade for the first two days of mechanical ventilation and received double the mean daily and cumulative doses of opioids compared to patients with non-COVID-19 related acute respiratory failure. The decreased use of HFOV in the COVID-19 cohort may be related to concern about environmental contamination; specifically, without a closed circuit or scavenging system, exhaled gases are released into the patient’s room. While the decreased use of HFOV could be related to a general trend in paediatric critical care management, there are no data to support such a speculation.

Static compliance (Crs) was measured on enrollment in our COVID-19 cohort. Early in the pandemic, it was suggested that adults with COVID-19 ARDS showed a different clinical phenotype than non-COVID-19 ARDS patients [23,24,25]. In small observational studies published early during the pandemic, it was observed that the severity of hypoxemia was out of proportion to impairment in respiratory system mechanics with patients having a relatively well-preserved respiratory system compliance [26,27,28,29,30]. However, in later studies, these observations were not confirmed, and, in fact, the relationship between Crs and ARDS severity quantified by the PaO2/FiO2 ratio was found to be similar between COVID-19 and non-COVID-19 ARDS patients [29, 30]. Such observations have not been reported before in children. Our study is the first to report lung mechanics in COVID-19 paediatric acute respiratory failure, and we found Crs normalized to bodyweight (Crs/kg) to be compatible with values previously reported for moderate PARDS [31].

Interestingly, those patients in the COVID-19 cohort were more likely to have abnormal neurocognitive functioning at baseline as assessed by PCPC and POPC scoring. The explanation for this association is unknown, but it may be speculated that neurologic sequelae of COVID-19 infection may have been a factor. At follow-up per parent report, cognitive and functional impairment increased to 74% in the COVID-19 cohort, and just over half of the children had a worsening of cognitive and functional outcome at follow-up as compared to pre-COVID-19 illness baseline. As the study was not designed or powered to assess long-term outcomes, we can only speculate as to the aetiology of the worsening of cognitive and functional outcome after discharge. While the number of patients with follow-up (n = 19) is limited, their follow-up cognitive and functional outcome and quality of life scores are low as compared to a healthy paediatric population.

Our study demonstrates that obesity in mechanically ventilated paediatric patients with SARS-CoV-2 infection does not affect outcomes which is consistent with the similarly matched cohort without COVID-19. However, our study is not designed or powered to assess a causal relationship between obesity and outcome. Variable findings have been reported regarding the relationship of obesity, severity of COVID-19 infections, and outcome with inconsistent explanations [25,26,27, 31,32,33,34]. Interestingly, Bouzid and colleagues noted that obesity was less common in adult patients presenting to the emergency department with the omicron variant as compared to those presenting during the delta variant wave [34].

The increased representation of the Hispanic/Latino population in the COVID-19 cohort is noteworthy. It is unclear whether this finding is related to a genetic predisposition, pattern of spread of COVID-19 in the communities served by the enrolling PICUs, or baseline differences in the PICUs included in the two cohorts analysed. Of note, there was no difference in race and ethnicity by obesity status. Our limited sample size precludes additional analysis or speculation.

The finding of increased use of rescue ECMO is not surprising given the outcome differences between the two cohorts. However, this difference may be related to changes in clinical practice patterns over time (2020–2022 for COVID-19 vs 2009–2013 for RESTORE). Additionally, the severity of the various COVID-19 variants may have played a role in the need for ECMO cannulation at various points of the pandemic.

While we report the largest prospective cohort of paediatric patients with severe COVID-19 infection requiring mechanical ventilation across international sites, our study has limitations. The number of COVID-19 patients with follow-up data is particularly low as contact information could not be obtained for parents who could not visit the hospital because they were COVID-19 positive and/or due to restrictive visitation policies. For those with contact information, the non-response rate was high. An uncontrolled confounding variable is the various strains of SARS-CoV-2 virus that circulated throughout the pandemic. Also, natural and vaccine-induced immunity has changed the picture for COVID-19 infection in adults and children. It must be noted that the matched cohort design can only yield associations, not causative conclusions. Furthermore, the incidence and potentially implications of obesity may have varied pre- and post-pandemic. Lastly, the period of data collection for the two studies (2020–2022 vs. 2009–2013) may reflect different clinical practice patterns, for example, use of pre- and post-intubation noninvasive respiratory support, administration of neuromuscular blockade, and use of HFOV, rather than intrinsic differences in patients with and without COVID-19 infection. However, PALICC-2 and PARDIE publications note that overall PARDS management has not changed in a clinically significant manner during this timeframe [12, 16, 17].

Conclusion

Children with acute respiratory failure due to COVID-19, compared with matched controls without COVID-19 infection, were more likely to be obese, Hispanic/Latino, and cognitively or functionally impaired at baseline. More children with COVID-19 also had pre-existing respiratory and neurologic conditions. While the risk of mortality was lower in the COVID-19 cohort, observed mortality was not different between groups. However, children with COVID-19 experienced fewer ventilator-free days and, among survivors, longer PICU stays.

Children with COVID-19 related acute respiratory failure suffered disproportionately from chronic conditions and required more critical care support than children with acute respiratory failure without SARS-CoV-2 related infection. Mid- and long-term outcome for children with acute respiratory failure related to COVID-19 infection requires continued investigation.