Baseline characteristics, management, and outcomes of 55,270 children and adolescents diagnosed with COVID-19 and 1,952,693 with influenza in France, Germany, Spain, South Korea and the United States: an international network cohort study

Study design, setting and data sources

This study is part of the Characterizing Health Associated Risks, and Your Baseline Disease In SARS-COV-2 (CHARYBDIS) study, a large-scale multinational cohort study using routinely-collected primary care and hospital EHRs, hospital billing data and insurance claims data from the US, Europe (the Netherlands, Spain, the UK, Germany and France) and Asia (South Korea and China).

From the nineteen databases contributing data to CHARYBDIS, only those with data on patients below the age of 18 years with a clinical diagnosis of COVID-19 or a SARS-CoV-2 positive test between January and June 2020 were included. A cohort of children/adolescents diagnosed with seasonal influenza in 2017-2018 was included for comparison.

To be included in the study, databases had to have a minimum of 140 children/adolescents (aged below 18 years). This cut-off was deemed necessary to estimate with sufficient precision (confidence interval width of ±5%) the prevalence of a previous condition or 30-day risk of an outcome affecting 10% of the study population. Data results for this paper were extracted from CHARYBDIS results on the 22^nd of September 2020. Figure 1 represents the selection process of the databases for this study. Eleven databases fulfilled the inclusion criteria: STARR-OMOP (US)[12], CU-AMC HDC (US), HealthVerity (US), CUIMC (US), OPTUM-EHR (US), Premier (US), IQVIA-OpenClaims (US), IQVIA-LPD France (France), IQVIA-DA Germany (Germany), SIDIAP (Spain)[13], HIRA (South Korea). Among these, five databases contributed to the hospitalized cohort: Premier, OPTUM-EHR, IQVIA-OpenClaims, CUIMC, HIRA. A more detailed description of the included data sources is available in Supplementary Table 1.

• Download figure
• Open in new tab

Figure 1. Database selection process

Colorado University Anschuz Medical Campus Health Data Compass (CU-AMC HDC), Columbia University Irving Medical Center (CUIMC), Clinical Practice Research Datalink (CPRD), Data Analyzer (DA), Daegu Catholic University Medical Center (DCMC), Health Insurance Review & Assessment Service (HIRA), Integrated Primary Care Information (IPCI), Longitudinal Patient Data (LPD), Nanfang Hospital COVID-19 Research Database (NFHCRD), Information System for Research in Primary Care (SIDIAP), STAnford medicine Research data Repository (STARR-OMOP), Tufts Research Data Warehouse (TRDW), Department of Veterans Affairs (VA-OMOP)

*No prior observation time available and therefore excluded from description of comorbidities.

Study participants and follow-up

Two non-mutually exclusive cohorts were included: 1) children/adolescents with COVID-19 diagnosis or a SARS-CoV-2 positive test (index date was the first of the two events), 2) children/adolescents hospitalized (index date was hospitalization date), with COVID-19 diagnosis or a SARS-CoV-2 positive test 21 days before or after hospitalization date. A similar diagnosed cohort of children/adolescents with seasonal influenza diagnosis or positive influenza test in 2017-2018 was also included. Study participants could contribute information to all cohorts (diagnosed or hospitalized with COVID-19, or diagnosed with influenza). Individuals diagnosed with COVID-19 could be part of the hospitalized cohort if COVID-19 was diagnosed during or at the time or 21 days before hospitalization. Cohort participants were followed in each cohort from the index date to the earliest of death, end of the observation period, or 30 days after. The codes used to identify COVID-19 are described in Supplementary Table 2. In order to describe history of comorbidities, only participants with at least 365 days of prior observation before index date were included. Children below age one were excluded from the cohorts requiring 365 days of prior observation.

Baseline characteristics, symptoms, drug use and outcomes of interest

Baseline information on age at index date and conditions up to 1 year before index date were identified. Conditions were ascertained based on the Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT) hierarchy, with all descendant codes included. Detailed definitions of each condition can be consulted in Supplementary Table 2.

Symptoms recorded at index date included fever, cough, dyspnea, malaise or fatigue, myalgia, anosmia or hyposmia or dysgeusia, gastrointestinal tract symptoms, diarrhea, vomits and nausea. All drugs prescribed/dispensed during the 30-days follow-up after the index date were ascertained. Individual medications were categorized using Anatomical Therapeutic Chemical (ATC) groupings. For the study of medications potentially used for COVID-19, we assessed all medications included in at least two randomized controlled trials [14]. The list was further enriched with suggestions from key stakeholders including regulatory agencies, key opinion leaders, and pharma industry. Medicines of interest were grouped into: 1) repurposed medications- those with alternative indications but believed to be efficacious as antivirals, 2) adjuvant therapies - used allegedly for the treatment or prevention of COVID-19 complications [15]. All conditions and medications and additional time windows (a month prior and on index date) are reported in full and are available in an interactive website: https://data.ohdsi.org/Covid19CharacterizationCharybdis/.

The 30-day outcomes described in the diagnosed cohorts included hospitalization, death, pneumonia, and multi-system inflammatory syndrome in children (MIS-C, Kawasaki disease, or toxic shock syndrome). In the hospitalized cohorts, we additionally report sepsis, total cardiovascular disease events, acute respiratory disease syndrome (ARDS), cardiac arrhythmia, and bleeding. The definition of each outcome is provided in Supplementary Table 2.

Statistical analyses

All data were standardized to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM).[16] A common analytical code for the CHARYBDIS study was developed for the OHDSI Methods library which was run locally in each database. Only aggregate results from each database were publicly shared. The CHARYBDIS protocol and source code can be found at https://github.com/ohdsi-studies/Covid19CharacterizationCharybdis. Results were extracted from CHARYBDIS on October 1st 2020.

Demographics, history of comorbidities, symptoms and outcomes were summarized as proportions, calculated by dividing the number of people within a given category by the total number of people. The proportion of users of each medication was determined for the hospitalized children/adolescents as the percentage of subjects who had >=1 day during the time window overlapping with a drug use period for each medication or drug class of interest. Utilization of repurposed and adjuvant drugs up to 30 days after admission was described.

The distribution of conditions a year prior to index, symptoms at index, outcomes and medications up to 30 days post-index date in COVID-19 diagnosed cohort were compared to the hospitalized COVID-19 or the diagnosed influenza cohorts. Standardized mean differences (SMD) were calculated when comparing the characteristics of study cohorts.

We performed a sensitivity analysis describing characteristics of cases with no prior observation time in order to understand the impact of lack of prior observation time in the results.

We used R version 3.6 (R Foundation for Statistical Computing, Vienna, Austria) for data visualization. All the data partners obtained Institutional Review Board (IRB) approval or exemption to conduct this study, as required.

Demographics

Age at diagnosis of COVID-19 varied across regions. In SIDIAP-Spain, CUIMC-US and STARR-US the majority of children with COVID-19 were diagnosed at ages 0 to 4 years (around one third), while the proportion of children from 0 to 4 years was only 11.4% and 11.6% in IQVIA-LPD-France and CUIMC-US, respectively. More consistently, most of the hospital admissions were seen in the younger groups (0-4 years), e.g. 40.6% of those hospitalized in IQVIA-OpenClaims-US, and 76.2% in Premier-US. Male gender was more common in all databases except for IQVIA-LPD-France (46.8% male) and STARR-US (49.2%) (Table 1).

Comorbidities

Assessed in the year before index date, asthma was the most common baseline comorbidity in patients diagnosed with COVID-19, affecting 10.1% (SIDIAP-Spain) to 33.6% (IQVIA-OpenClaims-US), followed by obesity (from 1.9% in IQVIA-LPD-France to 19.4% in OPTUM-EHR-US). We observed a high prevalence of congenital malformation(s) (7.9% of those diagnosed in IQVIA-OpenClaims-US to 10.8% in SIDIAP-Spain), neurodevelopmental disorders (1.0% of those diagnosed in IQVIA-LPD-France to 8.8% in IQVIA-OpenClaims-US), heart disease (1.2% in IQVIA-LPD-France to 13.1% in IQVIA-OpenClaims-US), type 1 diabetes mellitus (0.2% in SIDIAP-Spain to 0.5% in IQVIA-OpenClaims-US), cancer (0.3% in SIDIAP-Spain to 4% in OPTUM-EHR-US), and chromosomal disorder(s) (0.4% in SIDIAP-Spain to 1.3% in IQVIA-OpenClaims-US). All of these were more common amongst hospitalized children/adolescents with COVID-19 as compared to the diagnosed with COVID-19 cohort (SMD>0.1): 38.8% asthma, 13.1% obesity, 37.0% heart disease (5.1% congenital heart disease), 13.2% cancer, 5.8% chromosomal disorder(s), 22.7% congenital malformation(s), and 4.1% prematurity in IQVIA-OpenClaims-US (Figure 2).

• Download figure
• Open in new tab

Figure 2. Prevalence of previous comorbidities among children/adolescents (<18 years of age) diagnosed (X axis) compared to hospitalized (Y axis) with COVID-19. SMD = Standardized Mean Differences in prevalence between X and Y.

Symptoms

Figure 3a shows recorded symptoms at index date for diagnosed vs hospitalized COVID-19 patients while Figure 3b displays symptoms of COVID-19 vs influenza cohorts. The most common reported symptom in COVID-19 was fever, seen in 4.8% (SIDIAP-Spain) to 26.4% (CUIMC-US) of diagnosed cases, and higher (up to 28.1% in CUIMC-US) among hospitalized cases. Second most common was cough, recorded in 4.7% (SIDIAP-Spain) to 14.5% (Premier-US) amongst the diagnosed, and lower (e.g. 2.8% in Premier-US) in hospitalized children/adolescents. Bronchiolitis was recorded in 0.5% (SIDIAP-Spain) to 3.1% (IQVIA-OpenClaims-US) of the diagnosed, and higher (up to 11.9% in Optum EHR-US) in the hospitalized. Gastrointestinal tract symptoms were also common at index date, recorded in 0.5% (HealthVerity-US) to 12.5% (SIDIAP-Spain) in diagnosed, and up to 15.0% (Optum-EHR-US) among hospitalized. Anosmia was <=1% in all participating databases, both in diagnosed and admitted COVID-19 patients. Compared to influenza, COVID-19 diagnosed children/adolescents had less frequently reported symptoms in most databases, with the only exceptions of: dyspnea (e.g. 4.1% in COVID-19 vs 0.4% in influenza in IQVIA-OpenClaims-US), bronchiolitis (e.g. 3.1% in COVID-19 vs 0.2% in influenza in IQVIA-OpenClaims-US), anosmia/hyposmia/dysgeusia (e.g. 0.5% in COVID-19 vs 0.0% in influenza in IQVIA-OpenClaims-US), and gastrointestinal tract symptoms (e.g. 4.9% in COVID-19 vs 3.4% in influenza in IQVIA-OpenClaims-US).

• Download figure
• Open in new tab

Figure 3. Symptoms recorded at index date among children/adolescents (<18 years of age)

In-hospital treatments

Use of drugs during hospital admission for COVID-19 amongst children/adolescents is reported in Figures 4a (repurposed) and 4b (adjunctive therapies), based on data from CUIMC-US, OPTUM-EHR-US, Premier-US, and HIRA-South Korea. Repurposed treatments were not commonly used (<10% in all databases), with lopinavir-ritonavir used in <6% in HIRA-South Korea but not in US databases, azithromycin from 4% to 6% in all databases, hydroxychloroquine in 1.5% (Premier-US) to just above 4% (OPTUM-EHR-US), and oseltamivir only in the US, from <1% in Premier-US to just over 5% in CUIMC-US.

• Download figure
• Open in new tab

Figure 4. 30-day in-hospital use of treatments among children/adolescents (<18 years of age) with COVID-19

Adjunctive therapies were more common, with systemic corticosteroids used in 6.8% (HIRA-South Korea) to 37.6% (CUIMC-US). Famotidine was the second most common adjunctive treatment, used in 9.0% (OPTUM-EHR-US) to 28.1% (CUIMC-US). Concomitant antithrombotic therapy was also common in the US but not in HIRA-South Korea (no use reported), including aspirin (2.0% in Premier-US to 21.4% in CUIMC-US), heparin (2.2% in Premier-US to 18.1% in CUIMC-US), and enoxaparin (2.8% in Premier-US to 14.8% in CUIMC-US). Antibiotics (ceftriaxone, amoxicillin, fluoroquinolones), vitamin (D and C) supplements, and immunoglobulins were also used with high variability between the contributing databases.

Health outcomes

Outcomes in the 30-day period following the diagnosis of COVID-19 and hospitalization with COVID-19 are summarized in Table 1 and Figure 5a. Hospitalization was observed in a low proportion (0.3% in HealthVerity-US, 1.4% in SIDIAP-Spain) in the ambulatory/claims databases with testing and testing results data available; in 13.2% in IQVIA-OpenClaims-US; and more frequent (7.1% in OPTUM-EHR-US to 36.1% in Premier-US) in hospital EHR databases. Pneumonia was the most common complication, diagnosed in a wide range from 0.1% (HealthVerity-US) to 6.5% (IQVIA-OpenClaims-US) of those diagnosed, and in 6.8% (HIRA-South Korea) to 20.3% (IQVIA-OpenClaims-US) of those hospitalized with COVID-19. ARDS was the most common in-hospital outcome, affecting 8.1% (CUIMC-US) to 22.7% (Premier-US) of those hospitalized with COVID-19. Sepsis during admission was observed in <2% (N<5; HIRA-South Korea) to 9.4% in Premier-US. Other less common outcomes are reported in Table 1. MIS-C was rare, and seen only in <0.1% (N<5; SIDIAP-Spain) to 0.3% (IQVIA-OpenClaims-US).

• Download figure
• Open in new tab

Figure 5. Main 30-day outcomes among children/adolescents (<18 years of age)

A comparison of outcomes in those diagnosed with COVID-19 vs those diagnosed with influenza in previous years is depicted in Figure 5b. Hospitalization rates were higher for COVID-19 vs influenza-diagnosed children/adolescents (e.g. 13.2% vs 0.9% in IQVIA-OpenClaims-US, 33.2% vs 7.4% in CUIMC-US, 30.8% vs 3.7% in STARR-US). Similarly, pneumonia was more common in COVID19-diagnosed participants (e.g. 3.1% vs 0.7% in SIDIAP-Spain, 6.5% vs 2.4% in IQVIA-OpenClaims-US); and so was ARDS (e.g. 2.5% vs 0.1% in IQVIA-OpenClaims-US, 3.3% vs 0.7% in CUIMC-US) and - despite rare-MIS-C (e.g. 0.3% vs 0.0% in IQVIA-OpenClaims-US, 3.1% vs <0.2% in CUIMC-US). Other outcomes had more similar risks in both viral infections (see Figure 5b and Supplementary Table 3).

In a sensitivity analysis, we replicated the analyses including participants who had no prior history available in their EHRs (Supplementary Table 4). Differences in symptoms or outcomes were modest, however it demonstrated the expected incompleteness in prevalent comorbidity.

Comparison with other studies

The proportion of children/adolescents under 18 years of age from all observed COVID-19 cases in each database varied from 2.6% in CU-AMC HDC-US to 7.1% in IQVIA-DA-Germany. These values are higher than what has been previously reported in other studies from China (2%) [17], Spain (2%) [18], the US (1.7%) [5], or the UK (0.9%) [19], but in line with the WHO dashboard which reported that children below the age of 14 years old accounted for 3.7% of all COVID-19 cases [1], or the Australian Health Protection Agency which has reported that children/adolescents below the age of 19 years accounted for 4% of confirmed COVID-19 cases in Australia.[20]

Asthma and obesity were the most common baseline comorbidities in children/adolescents with COVID-19; this was expected, and is in keeping with disease prevalence among a general pediatric population.[21] More strikingly, we observed a high prevalence of conditions that are relatively rare in children/adolescents, including congenital malformation/s, neurodevelopmental disorders, heart disease, type 1 diabetes mellitus, cancer, and chromosomal disorder/s. These conditions were more frequent amongst hospitalized children/adolescents with COVID-19 than the overall cohort of those diagnosed with COVID-19. This is in line with previous studies suggesting that children with comorbidity history have higher risk of critical care admission [19,22].

Despite the great variability in the reporting of symptoms across databases, we found that the most commonly identified symptoms in children/adolescents with COVID-19 were consistently fever, cough, dyspnea, and gastrointestinal tract symptoms which are in line with previous studies [7,19,22]. However, the frequency of reported symptoms in our study is generally lower than what has been previously reported in the pediatric literature [7,19,22], suggesting an underestimation in the register of symptoms in the form of structured data in busy actual care settings. An important finding is that COVID-19 diagnosed children/adolescents presented with higher rates of dyspnea, anosmia, and gastrointestinal symptoms than children with seasonal influenza. This information is clinically relevant for differential diagnosis between COVID-19 and influenza among children/adolescents.

We observed great heterogeneity across countries in the use of in-hospital treatments among children/adolescents with COVID-19, which is in line with previous studies in adults [23]. Our analysis suggests little use of antiviral therapies overall, with about 5% of children/adolescents hospitalized with COVID-19 using lopinavir/ritonavir in South Korea (but none in the US); a variable proportion of use of oseltamivir (1% to 5%) in the US (but none in South Korea); a 4-5% of use of azithromycin, and limited use of hydroxychloroquine, ranging from 1.5% to around 4% in both countries. These values are lower than what we previously reported in adults (e.g. use of hydroxychloroquine was 57.9% in the overall population vs. 1.5% in children/adolescents in Premier-US) [23], but in line with recent European cohort studies in hospitalized children/adolescents with COVID-19. A study in the UK reported that 6% (38/591) of hospitalized children/adolescents with COVID-19 received antiviral drugs (30 received acyclovir, 7 received remdesivir, and 3 received chloroquine or hydroxychloroquine) [19], while a study in 25 European countries including 582 children/adolescents found that 7% were treated with hydroxychloroquine, 3% with remdesivir, 1% with lopinavir–ritonavir, 1% with oseltamivir.[7] In contrast, we observed substantial use of different adjunctive therapies; corticosteroids were used in 25-35% in the US but 7% in South Korea. Famotidine (2 to 20%), aspirin (10 to 30%) and vitamin D (2 to 15%) were used in the US but not in South Korea. Antibiotics were also commonly used, with ceftriaxone and amoxicillin amongst the most commonly prescribed. This is consistent with the previous study in in UK where antibiotics were given to 69% (415/601) of hospitalized children with COVID-19.[19]

It is reassuring that occurrence of severe outcomes during the 30 days after diagnosis of COVID-19 was rare in our study, which is in line with previous studies [7,18,19,22]. MIS-C was relatively uncommon, affecting 0.5%-3.1% of all diagnosed cases, but up to 0.9%-7.6% of those hospitalized with COVID-19. These results are in line with previous studies from Europe and the US which have suggested that COVID-19 may be associated with MIS-C in children.[24-26] A separate cohort study found recently that 11% of children with COVID-19 admitted to hospitals in the UK developed MIS-C.[19] These findings are of special relevance given the severity of this condition. Our results also confirmed MIS-C might be specifically related to COVID-19, since it was much less common in patients hospitalized with influenza. Overall, all outcomes were more frequent in children/adolescents with COVID-19 diagnosis than those with a diagnosis of seasonal influenza in 2017-2018, suggesting more severe disease prognosis in children with COVID-19 than influenza. Future research is needed to characterize and determine the long-term outcomes of children/adolescents affected with COVID-19.

Strengths and limitations

This study has some limitations. First, this study is descriptive in nature. The observed differences between groups should therefore not be interpreted as causal effects. Second, our results are based on secondary data from electronic records collected for administrative and clinical management purposes and re-used for research which may affect the completeness of data recorded and may have erroneous entries, leading to potential misclassification. Such incompleteness could be differential in some instances (e.g. hospital vs primary care settings) and risk information bias for the proposed comparisons. Moreover, variability in coding practices and the heterogeneity in healthcare settings inherent to this study may account for some of the observed variability in the prevalence of comorbidities, symptoms and outcomes. Hospital admissions are also variably recorded in the analyzed databases. The under-reporting of symptoms observed in these data is a key finding of this study, and should be taken into consideration in previous and future similar reports from ‘real-world’ cohorts. Our analysis of mortality is similarly subject to differences by database. For example, data on inpatient deaths are more complete while primary care or outpatient death events are typically imported into a given database from a national or local death register. Finally, the currently available data is restricted to the first six months of 2020 coinciding with the peak of the COVID-19 peak in the studied countries. As data accumulates over time future updates of the results will provide the opportunity to study more recent cohorts of COVID-19 patients, who seem to have a better prognosis overall compared to those diagnosed in the first half of the year.

This study is unique as our approach to characterizing children with an international scope allows for a wide range of variation in healthcare systems and policies against the COVID-19 pandemic. This enables a more complete understanding of the implications of the pandemic for different countries and regions in scope of an international comparison. It also poses a layer of heterogeneity that needs to be considered in the interpretation of our findings, opening a window for new research questions that need to be addressed; particularly around the public health approach for controlling the pandemic spread and severity in children/adolescents and overall. This study represents, to our knowledge, the largest cohort study on pediatric COVID-19 to date, and the only study to provide a comparison with children/adolescents with seasonal influenza in 2017-2018. Our data confirm low rates of complications in children with COVID-19, with severe cases clustering amongst children with previous comorbidities. Despite the large sample size available, MIS-C appears rare.