Volume and Severity of Pediatric COVID-19 Hospitalizations in the United States Free

We extracted data from pediatric tertiary care PHIS hospitals for patients hospitalized with COVID-19 (identified by primary or secondary diagnosis code; International Classification of Diseases 10th Revision, ICD-10 code: U07.1) admitted from April 5, 2020 to August 27, 2022 and discharged by September 30, 2022. Our analysis was restricted to patients less than 12 years old at admission and excluded neonates (<30 days). Hospitals submitting complete data during the full study period were included (N = 45 of 50 hospitals). This study used administrative data from PHIS that is considered nonhuman subjects research and classified as exempt by our hospital’s Institutional Review Board.

Variables collected for each admission included length of stay, age at admission, gender, race, ethnicity, insurance payor, and principal ICD-10 diagnosis code. Using flags available in PHIS based on diagnosis, procedure, and billing codes, we identified admissions with any ICU utilization, admitted from the emergency department, requiring mechanical ventilation or extracorporeal membrane oxygenation during admission, and in-hospital mortality. Principal diagnosis was grouped into the following primary diagnoses: COVID-19 (ICD-10 diagnosis code: U07.1 – COVID-19), acute respiratory failure (J96.0 – acute respiratory failures; J80 – acute respiratory distress syndrome; R06.03 – acute respiratory distress), multisystem inflammatory syndrome (M35.81 – multisystem inflammatory syndrome), COVID pneumonia (J12.82 – pneumonia caused by coronavirus disease 2019), and all other principal diagnoses. For those with another principal diagnosis, we used PHIS flags based on ICD-10 principal procedure codes to identify surgical procedures.

We examined weekly trends in volume of COVID hospitalizations overall and by ICU utilization. Consistent with prior studies in both adults and children, we considered ICU utilization as a suitable measure for severe disease.^7,12,13  Our primary dependent variable was the ratio of the number of COVID hospitalizations requiring versus not requiring ICU-level care. In addition, we examined the ratio of hospitalizations with COVID as the primary diagnosis versus secondary diagnosis to describe trends in hospitalizations for patients admitted primarily for COVID versus other admissions with a COVID diagnosis.

We describe sociodemographic and clinical characteristics using median (interquartile range [IQR]) for continuous variables and frequency (percent) for categorical variables. Using unadjusted time series analysis, we estimated the weekly trend in the ratio of COVID hospitalizations requiring versus not requiring ICU care. The ratio of ICU to non-ICU admissions was log-transformed before analysis. Model estimates were exponentiated and reported as the annualized percentage change with the 95% confidence interval (CI). Pearson correlation was used to examine the association between total COVID admission volume and the ratio of ICU to non-ICU admissions. We also examined the weekly trend in the ratio of ICU to non-ICU admissions in a time series analysis adjusting for total COVID volume. We also explored the trend in the ratio of admissions with COVID as the primary versus secondary diagnosis using a similar time series analysis. All analyses were performed in SAS (v9.4; Cary, NC) and P < .05 was considered statistically significant.

There were 38 160 hospitalizations with COVID-19 across 45 PHIS hospitals during the study period (per hospital, median = 763, range 272–2438). Sociodemographic and clinical characteristics are presented in Table 1. Median age was 2.4 years (IQR = 0.7–6.6). The majority of patients were male (56.1%), white race (55.8%), non-Hispanic (68.3%), with public insurance (62.7%). Median length of stay was 2.0 days (IQR = 1.0–4.0), 88.5% stayed 7 days or fewer, whereas 61.6% stayed fewer than 3 days. Overall, 18.9% required ICU-level care and 7.0% mechanical ventilation. In-hospital mortality was rare (0.6%). COVID-19 was the most common primary diagnosis (53.8%), although less than 1% had another COVID-related diagnosis, and the remaining 45.3% were classified as other diagnoses. Among those with another primary diagnosis, the most common diagnoses were: seizures, epilepsy, or convulsions (8.1%); sepsis (6.1%); appendicitis (5.7%); bone fracture (4.4%); diabetes (3.4%); lower respiratory diagnoses including bronchiolitis, acute bronchitis, and non-COVID pneumonia (3.1%); mental health conditions (3.0%); and asthma (2.2%). In addition, 21.1% of those with another primary diagnosis underwent a surgical procedure during admission, and among those, the most common procedure was appendectomy (16.5%).

The weekly trend in COVID hospitalizations overall and by ICU utilization is presented in Fig 1A. Overall admission volume is driven primarily by those without ICU-level care. The weekly trend in ratio of ICU to non-ICU admissions is presented in Fig 1B. Our analysis indicated that this ratio declined 14.5% annually (95% confidence interval [CI]: −21.7% to −7.26%) or equivalently 1.21% monthly (95% CI: −1.81% to −0.60%; P < .001), from a ratio of 0.47 ICU to non-ICU at the beginning of the study in April 2020 to 0.20 at the end of the study in August 2022. Total volume of COVID hospitalizations was inversely associated with the ratio of ICU to non-ICU admissions (correlation coefficient = −0.26; P = .003), indicating as total COVID volume increased, the ratio of ICU to non-ICU decreased. Results for the trend were similar after adjusting for total COVID volume (−13.4% annually; 95% CI: −21.1% to −5.72%; or −1.12% monthly; 95% CI: −1.76 to −0.48; P < .001) and there was no association between total COVID volume and the ratio of ICU to non-ICU admissions (−0.01% per each additional admission; P = .37).

The weekly trend in hospitalizations by COVID diagnosis hierarchy and ratio of primary to secondary COVID diagnosis are presented in Fig 2. Through the first year of the pandemic, admission volumes were evenly split between those with COVID as the primary versus secondary diagnosis, whereas in the second year and beyond, admissions with a primary diagnosis were more common. The ratio of admissions with primary versus secondary COVID diagnosis was 1.0 in April 2020 compared with 1.26 in August 2022. There was variation over time in the ratio (Fig 2B) and time series analysis indicated that the ratio was stable over time (+11.7% annually; 95% CI: −8.83 to 32.4; or +0.98% monthly; 95% CI: −0.74 to 2.70; P = .26). Results were similar adjusting for total COVID vol (+10.55% annually; 95% CI: −3.40% to 24.5%; or 0.88% monthly; 95% CI: −0.28% to 2.04%; P = .13).

There is concern over high rates of pediatric infection as successive waves of severe acute respiratory syndrome coronavirus disesase-2 variants have emerged since the start of the pandemic.^2–4,7  We considered data on pediatric infections from PHIS, a geographically diverse and comprehensive data source from tertiary care children’s hospitals across the United States. Periodic increases in pediatric hospitalizations with COVID-19 are evident, corresponding to similar increases in adult populations after introduction of new variants.^1–4  However, there is no evidence of corresponding increase in severity of illness, and the overall trend in severity among the pediatric hospital population is declining, although a large proportion of admissions were for primary diagnoses other than COVID-19.

We identified COVID-19 by primary or secondary diagnosis codes, thereby capturing both admissions related to COVID and incidental COVID among hospitalizations for other reasons. We were unable to discern admissions specifically for COVID-related symptoms given the limitations of billed diagnosis codes and administrative data. Nearly half of admissions had another primary diagnosis unrelated to COVID, likely representing a combination of both incidental COVID-positive hospitalizations for other reasons and patients admitted for COVID with another primary diagnosis, though we were unable to explore this further. We found the ratio of admissions with a primary versus secondary COVID diagnosis was variable but relatively stable over time. Future studies are needed to understand trends in incidental COVID admissions and any associations with hospital capacity, total population COVID volume, or other factors.

Our results are limited to hospitals in PHIS that are all large urban tertiary care pediatric hospitals and therefore, may not be generalizable to all pediatric COVID hospitalizations, in particular hospitalizations at community hospitals and/or hospitals in rural areas. Data are not included in PHIS until discharge and therefore, it is possible that the longest, most severe admissions and/or those admitted for diagnoses other than COVID were not captured as they had not been discharged by the end of our study period. However, the study period allowed for 1 month of follow-up time to capture discharges, thereby minimizing the number of missed admissions. Overall, only 2% of admissions were stays longer than 1 month and less than 0.5% were longer than 90 days. We examined the aggregate total number of admissions and did not include a population-level denominator such as the total patients tested for COVID-19 as this information is not available in PHIS. Availability of testing and test positivity rates have changed over time throughout the course of the pandemic, though we were unable to account for this and include only those admissions with a billed diagnosis for COVID-19. In addition, we were unable to account for vaccination status as this information is not available in PHIS. We considered the need for ICU-level care as a proxy for severe disease in the absence of additional clinical information to ascertain severity, such as symptoms. Importantly, there is no additional information available in PHIS to ascertain the reason for ICU admission that may be unrelated to COVID-19. However, this definition is consistent with prior studies among both adult and pediatric cohorts.^7,12,13 

Our results suggest that severity has remained relatively stable despite wide temporal variation in utilization, highlighting the important distinction at a population level between utilization and severe morbidity. Examining frequency of hospitalizations alone may not provide a complete picture of population-level disease.

There are many factors at the patient, provider, hospital, and system-wide levels that may affect volume, eg, likelihood to seek care, proclivity to admit (perhaps out of abundance of caution), and hospital capacity. Often, the most likely explanation for higher utilization is an underlying increase in rate of infection among the patient population. It is harder to provide a convincing explanation for the slight reduction in disease severity over time among hospitalized patients. Because the study period spans an era when vaccines were not widely available for young children, this observed decrease might be explained by improved clinical protocols, increasing levels of comfort caring for patients with COVID, and/or a moderation in virulence over successive waves of more transmissible viral variants. Given our study design, further explanation of our findings must remain speculative.

Our findings suggest that the risk of severe disease in children has not increased over the course of the pandemic, which may provide important context and a more nuanced and appropriate interpretation when pediatric hospitalizations are observed to increase. In addition, these findings may have important implications for understanding trends in severity for emerging variants of COVID-19 as well as other seasonal respiratory viruses, including influenza and respiratory syncytial virus. Future studies should similarly examine trends in severity and pediatric hospital utilization for COVID-19 and other seasonal respiratory viruses to inform public health and education policy.