Factors Associated With Severe Illness in Patients Aged <21 Years Hospitalized for COVID-19

We conducted a cross-sectional study with 6 children’s hospitals in the United States (in Arkansas, the District of Columbia, Florida, Illinois, Louisiana, and Texas). We abstracted charts of patients aged <21 years who were hospitalized with a positive SARS-CoV-2 nucleic acid amplification or antigen test or a provider diagnosis of COVID-19 at any point from July 1 to August 31, 2021, a period of high community transmission of SARS-CoV-2 in each hospital catchment area.²³ 

For each patient we abstracted demographic characteristics, medical history, presentation, hospital course, and outcomes. Using the provider documentation, abstractors determined hospitalization reason as either for COVID-19 or with an incidental positive SARS-CoV-2 test. Among those hospitalized for COVID-19, abstractors assigned COVID-19 as the primary reason for hospitalization or as a contributing reason (if provider documentation indicated that COVID-19 was partly responsible for the hospitalization, but not the clear primary reason). A patient with an incidental positive test was defined as having asymptomatic or mild infection and being hospitalized for a reason apparently unrelated to COVID-19. To audit the abstractors’ determinations and to minimize abstraction errors, 5% of all charts were reviewed by a second abstractor. When medical documentation was unclear, project leaders adjudicated hospital reason determination. Patients with multisystem inflammatory syndrome in children (n = 25) were excluded from this analysis and are described in Supplemental Table 5.

Among patients aged 2 to 20 years, patients were classified as having obesity if their BMI calculation or reported BMI was ≥95^th percentile, or if they had a clinical diagnosis of obesity. Other underlying medical conditions were abstracted based on provider documentation. Depending on number of organ systems affected, patients were categorized as having 0, 1, or ≥2 underlying medical conditions, restricting to 1 condition per organ system group.²²  Patients were considered fully vaccinated if they had received 2 doses of a messenger RNA–based COVID-19 vaccine ≥14 days before hospital admission date, consistent with completion of the primary series, as documented in the medical chart. During the study period, COVID-19 vaccination was recommended for persons aged 12 years and older, and the Pfizer-BioNTech was the only vaccine authorized for persons aged 12 to 17 years.²⁴  Prevalence of tachycardia and tachypnea was calculated using age-normalized cutoffs, as described elsewhere.^12,25  Chest radiographs were defined as either normal or abnormal, and abnormal radiographs were further classified as having an infiltrate or consolidation on the basis of interpretations provided in radiologist reports.

Viral coinfection was defined as having a positive viral test. Bacterial coinfection was defined as having a positive blood culture (not considered to be a contaminant), respiratory bacterial culture, or a provider-assigned diagnosis of bacterial infection, including bacterial pneumonia.

Highest level of respiratory support was defined preferentially as invasive mechanical ventilation (IMV) including extracorporeal membrane oxygenation, bilevel or continuous positive airway pressure (BiPAP or CPAP), high-flow nasal cannula (HFNC), or nasal cannula.

We defined “severe illness” as receiving HFNC, BiPAP, CPAP, or IMV at any point during the hospitalization to identify patients with significant respiratory illness. We did not include ICU admission or medication requirements in the definition of severe illness given possible interhospital differences in ICU admission parameters and critical care requirements for other disease processes, such as diabetic ketoacidosis.

We provided descriptive statistics for the full cohort, stratified by hospitalization reason. Among those hospitalized for COVID-19 (as either the primary or contributing reason for hospitalization), we further described clinical presentation and hospital course stratified by age groups (<1, 1 to 4, 5 to 11, 12 to 17, and 18 to 20 years).²⁶  We selected these age groups a priori based on COVID-19 vaccine eligibility cohorts, age-specific applicability of certain underlying medical conditions (eg, preterm birth, RSV infection, and obesity), and the goal of filling a literature gap in granular clinical information about infants and young children. Admission, peak, and/or nadir laboratory values were summarized. We conducted subgroup analyses among neonates (aged <1 month) and nonneonate infants (aged 1 to 12 months) and among patients aged ≥2 years with and without obesity, using 2-sided Mann–Whitney U and χ² tests, with significance established at α of .05.

We assessed factors associated with severe illness among patients hospitalized for COVID-19 using log-binomial mixed-effects regression models with hospital-specific random intercepts to account for correlations of observations from the same hospital. We used log-binomial rather than logistic or probit regression because the prevalence of severe illness was not rare, and prevalence risk ratios would provide more reasonable estimates of association than prevalence odds ratios. For each of the age groups (<1, 1 to 4, 5 to 11, 12 to 17, and 18 to 20 years), we selected exposure variables that are relevant to that group to determine the independent effect of each exposure on the prevalence of severe illness. We calculated the prevalence ratios (PR) with 95% confidence interval (CI). Because our goal was to help pediatric hospital providers recognize potential factors associated with severe illness in admitted patients, rather than to develop a composite risk incorporating other factors, we ran a distinct model for each exposure variable. Given the objective of this study, to raise awareness of potential factors for severe illness in children hospitalized with COVID-19, we did not correct for multiple comparisons. Although this raises the type I error risk, it lowers the greater risk of missing statistical associations that could prove valuable in clinical practice. All analyses were conducted in SAS (version 9.4; SAS Institute).

This activity was reviewed by the Centers for Disease Control and Prevention and the participating institutions and determined to be public health surveillance. It was conducted according to federal law and CDC policy (45 C.F.R. part 46.102(l)(2), 21 C.F.R. part 56, 42 U.S.C. Sect. 241(d), 5 U.S.C. Sect. 552a, and 44 U.S.C. Sect. 3501 et seq.).

Of 947 patients, 759 (80.1%) were hospitalized for COVID-19 and 188 (19.9%) had an incidental positive SARS-CoV-2 test. COVID-19 was the primary reason for hospitalization for 494 (52.2%) patients and a contributing reason for 265 (28.0%) patients. Two patients received a provider diagnosis of COVID-19 without documentation of a positive SARS-CoV-2 nucleic acid amplification or antigen test. In the full cohort, 491 (51.8%) were male, 205 (21.6%) were aged <1 year, 337 (35.6%) were aged 12 to 17 years, 266 (28.1%) were non-Hispanic Black, 282 (29.8%) were Hispanic, and 518 (54.7%) were insured through Medicaid. Of 394 vaccine-eligible patients in the full cohort (aged 12 to 20 years), 4 (1.0%) were fully vaccinated at admission. Of 318 vaccine-eligible patients hospitalized for COVID-19, 2 (0.6%) were fully vaccinated. Among those hospitalized for COVID-19 as the primary reason, 344 (69.6%) had at least 1 underlying medical condition and none were fully vaccinated (Table 1).

Presenting signs and symptoms among the 759 patients hospitalized for COVID-19 included objective fever (n = 503, 66.3%), shortness of breath (n = 389, 51.3%), and nausea or vomiting (n = 316, 41.6%). Tachypnea was present at admission in 176 patients aged 12 to 17 years (64.7%), and tachycardia was present in 28 patients aged 18 to 20 years (60.9%). “COVID-19” and “pneumonia” were the most common admitting diagnoses (Table 2). Median C-reactive protein values at admission exceeded normal values in all age groups (Table 3).

Among the 759 patients hospitalized for COVID-19, 287 (37.8%) had severe illness. Of all patients, 62 (8.2%) received IMV as the highest level of respiratory support (including 10 who received extracorporeal membrane oxygenation), for a median duration of 7 days; 9.6% (n = 26) of patients aged 12 to 17 years, and 13% (n = 6) of patients aged 18 to 20 years received IMV, for respective median durations of 9.5 and 17 days.

Of 573 patients hospitalized for COVID-19 with at least 1 chest radiograph, 79.8% had an abnormality, including 52.2% with an infiltrate or consolidation. Viral coinfection was identified in 101 patients aged <5 years (33.9%), compared to 12 patients aged 5 to 20 years (2.6%). Bacterial coinfection was identified in 50 patients aged <5 years (16.8%) and 87 patients aged 5 to 20 years (18.9%) (Table 2). Coinfection testing was common in patients with and without severe illness. Of those with severe illness, 62.4% had a viral coinfection test and 49.2% had a bacterial culture; respective findings among patients without severe illness were 49.1% and 40.0%.

A total of 227 (29.9%) patients were admitted to the ICU, for a median duration of 4 days. ICU admission prevalence ranged from 19.3% (n = 34) for patients aged <1 year to 39.7% (n = 108) for patients aged 12 to 17 years. Remdesivir use ranged from 12.5% (n = 22) for patients aged <1 year to 63.0% (n = 29) for patients aged 18 to 20 years. Baricitinib, Tocilizumab, and other monoclonal antibodies were used rarely and mostly among patients aged ≥12 years. Thirty-three (4.3%) patients were discharged on oxygen, of whom 9 had underlying chronic respiratory failure. Eleven (1.4%) died, 72.7% of whom were aged 12 to 17 years (Table 4).

Factors associated with severe illness in infants included RSV coinfection (PR = 3.64; 95% CI, 2.50–5.30), bacterial coinfection (PR = 1.88; 95% CI, 1.21–2.94), and presence of ≥2 underlying medical conditions versus none (PR = 1.82; 95% CI, 1.04–3.18). Preterm birth was not significantly associated with severe illness (PR = 1.45; 95% CI, 0.93–2.25) (Fig 1, Panel A).

Among patients aged 1 to 4 years, severe illness was associated with RSV coinfection (PR = 1.96; 95% CI, 1.31–2.93) (Fig 1, Panel B). Among 298 patients aged <5 years hospitalized for COVID-19, respiratory support was required among 71.3% of patients who had a viral coinfection and 38.1% of patients who did not have a viral coinfection (Supplemental Table 6).

Among patients aged 5 to 11 years, severe illness was associated with feeding tube dependence (PR = 2.77; 95% CI, 1.54–4.96), bacterial coinfection (PR = 2.34; 95% CI, 1.36–4.02), gastrointestinal condition (PR = 2.32; 95% CI, 1.26–4.27), and obesity (PR = 2.20; 95% CI, 1.25 –3.87) (Fig 1, Panel C).

Among patients aged 12 to 17 years, severe illness was associated with presence of 1 (PR = 2.95; 95% CI, 1.29–6.75) or ≥2 (PR = 3.19; 95% CI, 1.41–7.20) underlying medical conditions versus none, obesity (PR = 2.48; 95% CI, 1.74–3.55), feeding tube dependence (PR = 1.55; 95% CI, 1.08–2.23), and neurologic or neurodevelopmental disorder (PR = 1.40; 95% CI, 1.05–1.88) (Fig 1, Panel D).

Detailed findings are provided on patients hospitalized for COVID-19 as the primary reason (Supplemental Table 7) or as a contributing reason (Supplemental Table 8), or hospitalized with an incidental positive SARS-CoV-2 test (Supplemental Table 9).

Of 58 neonates hospitalized for COVID-19, 5 (9%) had a RSV coinfection, 6 (10%) were born prematurely, 8 (14%) required ICU admission, 7 (12%) had severe illness, and 1 (2%) died. The most common discharge diagnoses for neonates were “COVID-19” (n = 53, 91%), “neonatal fever” (n = 24, 41%), and “bronchiolitis” (n = 4, 7%). Of 118 nonneonate infants hospitalized for COVID-19, 37 (31%) had RSV, 31 (26%) were born prematurely, 26 (22%) required ICU admission, 53 (45%) had severe illness, and 1 (1%) died (Supplemental Table 10).

Among 260 patients aged 2 to 20 years with obesity hospitalized for COVID-19, 41.5% required ICU admission, 10.8% required IMV, and 1.5% died. Patients with obesity were more likely to present with tachypnea, tachycardia, and infiltrate or consolidation on chest radiograph (P < .001 for all), and they had higher admission C-reactive protein and ferritin (P < .001 for both) compared to patients without obesity (Supplemental Table 11).

In this multicenter retrospective study of pediatric patients hospitalized for COVID-19 during the Δ surge, we provide robust clinical details and age-stratified factors associated with severe illness related to COVID-19. Over one-third of patients in this cohort had severe illness requiring ventilatory support (HFNC, BiPAP, CPAP, or IMV). RSV coinfection in children aged <5 years and obesity in those aged 5 to 17 years were significantly associated with severe illness. Additional age-specific factors included neurologic or neurodevelopmental disorders, gastrointestinal conditions, feeding tube dependence, and bacterial coinfection. Overall, patients incidentally positive for SARS-CoV-2 comprised only one-fifth of admissions. Of the 394 (41.6%) vaccine-eligible patients, only 4 (1.0%) were fully vaccinated. These data illustrate the extent of clinical illness in a primarily unvaccinated pediatric population during the initial Δ variant surge in the United States.

Our findings approximate those from previous studies. Among patients aged <18 years hospitalized for COVID-19 in our cohort, 29.5% required ICU admission, 7.9% received IMV, and 1.5% died. These figures slightly exceed those in a pre-Δ study of patients aged <18 years with COVID-19–associated hospitalization in 14 states from March 2020 to May 2021, in which 24.8% were admitted to the ICU, 5.3% received IMV, and 0.5% died.¹⁴  In a Δ-era study from the same surveillance network from July 1 to December 18, 2021, 27.8% were admitted to the ICU, 6.3% received IMV, and 0.6% died.²¹  Discrepancies may be because of differences in the study populations, such as prevalence of underlying medical conditions, and because patients incidentally positive for SARS-CoV-2 (ie, admitted with a positive SARS-CoV-2 result but not for COVID-19) were included in the previous studies. Presenting signs and symptoms in our cohort were similar to those in pre-Δ studies, with a predominance of fever, cough, dyspnea, and nausea or vomiting.^10,29 

Consistent with previous studies of hospitalized patients,^17,30  we found elevations in inflammatory markers. Previous studies have found that inflammatory markers appear to be associated with intensive care admission^15,19  and organ dysfunction due to COVID-19.³¹  Remdesivir and glucocorticoids were frequently used in our cohort, which is consistent with recommendations from the National Institutes of Health and treatment authorizations during the study period by the United States Food and Drug Administration, and may indicate illness severity in this cohort.^32,33 

Few neonates hospitalized for COVID-19 in our cohort experienced severe illness, compared to nearly one-half of nonneonate infants and two-fifths of patients aged 1 to 4 years. This may be because of provider practice of hospitalizing neonates with relatively mild illness. It may also reflect the high prevalence of viral and bacterial coinfections among nonneonate infants and young children. Importantly, although these coinfections were associated with severe illness in our cohort, the causal pathologic relationship cannot be determined given the study design. Longitudinal studies are needed to understand the morbidity implications of concomitant COVID-19 and other illnesses. In light of our findings, this is especially important for coinfection with RSV and SARS-CoV-2 in young children.

Obesity had a strong association with severe illness in our cohort. Patients aged 5 to 11 and 12 to 17 years with obesity were over twice as likely to have severe illness compared to their peers who did not have obesity. This mirrors pre-Δ studies, which reported an association between obesity and requirements for IMV and critical care,^17,34  and between obesity and hypoxia.³⁵  Given this consistent association between obesity and COVID-19 severity, and the rising prevalence of childhood obesity in the United States,³⁶  pediatricians should follow national recommendations for obesity screening and referral³⁷  and promote COVID-19 vaccination for children and adolescents with obesity.

Strengths of this analysis are that we distinguished patients hospitalized with an incidentally positive SARS-CoV-2 test from those hospitalized for COVID-19, and we provided robust, age-stratified clinical data and regression analyses, which may assist health care providers with management decisions. The limitations of the study should also be considered. First, completeness and accuracy of data capture could not be guaranteed. We minimized abstraction errors and discrepancies through universal quality checks and random audits. Missing data may have contributed to underreporting of certain variables, resulting in sample sizes that were insufficient for age-specific prevalence ratio calculations. This limitation is especially relevant for vaccination, which may not be fully captured in inpatient records.³⁸  Second, our severe illness definition differs from previous studies, affecting comparisons.^2,14  We excluded ICU admission from our severity definition to control for differences in critical care requirement thresholds and to focus on severe respiratory illness as a more specific outcome measure. For some patients with respiratory coinfections, such as RSV, this definition may not exclusively reflect illness due to COVID-19. Third, because illness severity could influence coinfection testing and provider diagnoses, prevalence ratios for coinfections may be overestimated. However, since viral coinfection testing and bacterial cultures were conducted on patients with and without severe illness, confounding by indication should not dramatically alter findings. Fourth, distinguishing between an incidental positive test and COVID-19 as the reason for hospitalization was based on clinical judgement, which could introduce misclassification bias. Finally, given Omicron variant emergence,³⁹  the limited geographic range of our cohort, the fact that these data are derived from children’s hospitals, and increased pediatric COVID-19 vaccination eligibility and coverage since our study period closed, generalization of these findings may be limited.

During the July to August 2021 Δ variant surge, over one-third of patients aged <21 years hospitalized for COVID-19 in our cohort had severe illness. RSV was associated with increased prevalence of severe illness for patients aged <5 years. In those aged 5 to 17 years, obesity and the presence of ≥2 underlying medical conditions were associated with severe illness. The majority of vaccine-eligible patients were unvaccinated. These findings may help health care providers identify patients more susceptible to severe illness and determine appropriate management, support public health efforts to increase COVID-19 vaccination of persons aged ≥5 years, and inform discussions around vaccination for children aged <5 years. The findings also reveal a need for research on the severity and pathogenicity of respiratory viral coinfections with COVID-19. Finally, this analysis may offer a baseline against which the pathogenicity of Omicron and future variants could be compared.

The authors thank Anthony Bastiand, Troy Bienemy, Jerry Bridgham, Joyce Dalton, Laura Fisher, Barret Flagg, Jennifer Giovanni, Kara Hollis, Ashok Kurian, Amy Knight, Veena Nagarajan, Aimee Ossman, Emily Paganelli, Georgina Peacock, Nicole Pereira-Abara, Trescena Preacher, Andrea Romaniuk, Leila Sahni, Susan Stark, Sherry Sweek, Daniella Van Der Merwe, and James Versalovic.