The coronavirus disease (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus-2 has spread from a local outbreak in China to a global pandemic within months. On 31 December 2019, a cluster of cases with pneumonia of unknown cause emerged from Wuhan, China. On 30 January 2020, the WHO declared the coronavirus outbreak a Public Health Emergency of International Concern, and on 11 March 2020, a pandemic. As of 21 January 2021, there have been over 95.6 million confirmed COVID-19 cases and over 2.0 million associated deaths from 216 countries, areas or territories.1 Children under-19 years of age comprise a small proportion (1%–10%) of the total reported cases2–5 with a lower risk of developing critical illness from COVID-19 infection compared with adults.6 Prior systematic reviews of paediatric COVID-19 have described a mild disease in children with good outcomes.4 7 8 Since the publication of these reviews, the pandemic has spread extensively around the globe. In addition to pulmonary manifestations of COVID-19 in children, reports from Europe, North America, Latin America and Asia have emerged, describing a multisystem inflammatory syndrome children (MIS-C) related to COVID-19 infection.9–12 COVID-19 has also disrupted essential maternal and child health interventions, including outpatient visits and vaccinations for young children in most countries, further worsening the existing burden on healthcare provision and delivery.13
The objective of this review, in addition to providing a comprehensive update of the evolving paediatric COVID-19 literature, is a unique comparison of reported cases in low-income and middle-income countries (LMICs) to high-income countries (HICs) and of children with severe versus non-severe disease. Furthermore, the review provides a subgroup analysis of children presenting with symptoms of MIS-C and neonatal cases.
The protocol of the review is registered with PROSPERO (CRD42020183134). This systematic review is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA).
The review includes a comprehensive search of MEDLINE, Embase, WHO COVID-19 Database, Chinese COVID-19 Databases (CNKI and Wangfang), Latin-American and Caribbean Health Sciences Literature (LILACS) from 1 December 2019 to 8 January 2021. Complementary searches were conducted in Google Scholar, John Hopkins Health Resource, WHO news and the Chinese and US CDC Library. MedRxiv, BioRxiv and ChinaXiv were searched for preprints. No language restrictions were applied.
A search strategy was formulated and administered as shown in online supplemental table 1.
Observational studies reporting children (0–19 years old) with laboratory‐confirmed COVID-19 (serology or RT-PCR) were considered for inclusion. Studies with a subset of children 0–19 years were included if disaggregated data for children were provided. Studies were screened for any overlap in paediatric cases by reviewing institution details and the period reported. Review articles, case reports, commentaries and letters not presenting any original data were excluded. Case reports were excluded to reduce risk of selection bias and over-representation of extreme cases. Covidence Software (2016) was used for screening by two reviewers independently and in duplicate. Key reference lists were screened for additional studies.
Two reviewers conducted data extraction using a prepiloted data form. Data extracted included authors’ names, date of publication, study-design, city, country, number of cases, gender, comorbidities, travel and contact history, diagnostic tests for COVID-19, clinical details, laboratory tests, radiological findings, management and outcomes. Disaggregated data by age groups (0–5 years, 5–10 years and >10 years old) and reported disease severity was extracted where available. Criteria for severe disease were as defined within each individual study and included admission to intensive care units (ICUs), use of mechanical ventilation, multiorgan failure and presence of hypoxia (oxygen saturation <92%).
Individual study quality was evaluated independently by the review authors using quality assessment tools developed by the National Heart Lung and Brain Institute (NHLBI)14 (https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). Study quality was scored out of 8, based on clarity of study objectives, case definition, consecutive inclusion of cases, comparability of included patients, definition and measurement of outcomes, length of follow-up, statistical methods and results. Studies with score 6–8 were considered to be good quality, 4–5 considered fair quality and <4 considered poor quality.
Categorical data were summarised as counts and proportions. The pooled proportions of reported findings were calculated using Comprehensive Meta-Analysis 2.2.027 using random-effects model. I2 was calculated to examine statistical heterogeneity (I2>50% considered high heterogeneity). The clinical features and outcomes were compared according to (1) World Bank country income level (HICs versus LMICs)15 and (2) reported disease severity (severe versus non-severe) using pooled proportions and their 95% CIs, supplemented by relative-risk (RR). Subgroup analyses of children with MIS-C and neonatal cases were conducted.
After removal of duplicates, 3350 citations were screened for inclusion. Full texts of 198 studies were screened and 129 studies2 3 9–12 16–138 were included (online supplemental figure 1). Sixty-nine studies were excluded as they either presented overlapping data, did not provide age-disaggregated data for children or were commentaries, editorials or reviews. In terms of study setting, 13 studies were population-based national surveillance studies, 94 studies included only children admitted to hospital and 22 studies reported patients presenting to outpatient clinics or emergency departments (hospitalisation rate of 24.2%, 385/1590).
Sixty studies were from HICs (n=6528) and 69 studies from LMICs (n=3723). Almost one-third of included studies were from China (36/129, 28.0%),2 16–33 35 37–43 45 52 53 55–57 71 74 138 one-fifth were from the USA (24/129, 18.6%)3 9 10 47 58 60 64 68–70 79 81 82 85–87 93 98 109 110 115 120 133 136 together comprising almost half of the included sample size (n=4758, 46.4.%). The country of origin of included studies and study characteristics are summarised in online supplemental figure 2, tables 2 and 3.
Demographics and epidemiology
A total of 9335 children from the 129 case series were included in the meta-analysis. Of 8455 children for whom initial disposition was reported, 4851 were hospitalised (57.4%). Among them, 55.5% were men. The patient’s ages ranged from 0 to 17 years with mean age of 7.0±3.6 years. Ninety-one of the 129 studies reported age-disaggregated data for infection incidence as shown in online supplemental table 4. Nearly half of the cases were >10 years of age. Contact exposure to COVID-19 was reported in 64.0% of cases. Travel history to an epicentre was reported in 13.0% cases. At-least one underlying comorbidity was reported in 27.1% of cases. The most common reported comorbidities were immunosuppression (15.8%) and lung disease (12.5%).
Table 1 summarises the clinical manifestations reported in the studies. There were 13.1% asymptomatic cases (95% CI 10.4% to 16.3%) who presented primarily through contact exposure in family-clusters (parents, siblings and other relatives). The most common presenting symptoms were fever (63.3%, 95% CI 58.6% to 68.4%) and cough (33.7%, 95% CI 29.6% to 38.1%) followed by nausea or vomiting (20.0%, 95% CI 16.5% to 24.0%) and diarrhoea (19.6%, 95% CI 16.1% to 23.7%). Other symptoms included dyspnoea, nasal-symptoms, rashes, kawasaki-like symptoms, conjunctivitis, fatigue, abdominal pain and neurological symptoms. Sixty-seven of the 129 studies reported age disaggregated data for clinical features (online supplemental table 5). Clinical features were similar in the three age groups: ≤5 years, >5 to ≤10 years, >10 years with higher prevalence of abdominal symptoms in children>5 years.
Radiological and laboratory findings
One thousand five hundred and thirty cases out of 3670 (44.1%, 95% CI 39.5% to 48.9%) cases had radiological abnormalities; ground glass opacities (27.4%) were the most commonly reported abnormality.
Sixty-six studies provided details on laboratory-markers (table 2). Pooled analysis revealed increased C-Reactive Protein (CRP) (54.2%, 95% CI 41.5% to 66.3%), serum-ferritin (46.7%, 95% CI 32.3% to 61.7%), lactate dehydrogenase (LDH) (36.5%, 95% CI 26.5% to 47.8%) and d-dimers (35.2%, 95% CI 22.1% to 51.0%) as the most common abnormalities. Other reported abnormalities included elevated erythrocyte sedimentation rate (ESR), lymphopaenia, procalcitonin and biomarkers for organ injury including elevated levels of pro B-type natriuretic peptide, troponin and creatine kinase-MB as shown in table 2.
Details of clinical management are as shown in table 3. Commonly used therapies among hospitalised children were antimicrobials (32.2%, 95% CI 25.2% to 40.1%), intravenous immunoglobulin (IVIG) (19.5%, 95% CI 13.5% to 27.2%) and systemic-steroids (19.3, 95% CI 14.9% to 24.9%). Other treatment regimens included aspirin, inotropic drugs, inhaled interferon-α (IFN‐α), antimalarials and antivirals (ribavirin, oseltamivir, lopinavir, ritonavir and litonavir). Mechanical ventilation was provided to 490 patients (12.2%, 95% CI 9.7% to 15.3%).
Prognosis and severe cases
One thousand three hundred and fifty-nine patients (22.9%, 95% CI 17.6% to 29.2%) were admitted to ICUs (table 3). Thirty-eight studies provided disaggregated data for severe cases (table 4). A higher proportion of children with severe disease had symptoms consistent with MIS-C and received antimicrobials, inotropes and anti-inflammatory agents compared with those with non-severe disease. There were no deaths among children categorised as non-severe and 44 deaths among severe cases, where disaggregated data were provided. Hospital outcomes were reported for 9335 children; 8704 cases (88.9%, 95% CI 86.0% to 91.2%) were definitively discharged, 96 died and remaining children either remained hospitalised at the time of reporting or were readmitted.
Comparison of outcomes according to World Bank Country Classification
Of the included studies that reported clinical characteristics and outcomes, 60 studies were from HICs (n=6528) and 69 studies from LMICs (n=3723) as show in table 5. Studies in LMICs included a higher proportion of hospitalised children (1981/3723, 53.2%) compared with HIC studies (2897/6528, 44.4%). Abdominal symptoms and symptoms consistent with MIS-C were more frequently reported in HICs. A lower proportion of children in LMICs were admitted to the ICU (RR 0.56, 95% CI 0.50 to 0.63, p<0.05), mechanically ventilated (RR 0.32, 95% CI 0.26 to 0.39, p<0.05) and treated with different therapies; inotropes, antimicrobials, steroids, aspirin and IVIG. Only children in LMICs received inhaled IFN‐α. Among the hospitalised cases, 40 deaths were reported in HICs compared with 56 in LMICs (pooled proportion 2.9% vs 5.2%). Risk-adjusted mortality according to severity of illness could not be calculated due to lack of data (table 5).
Subgroup analyses of children presenting with MIS-C, and COVID-19 in neonates
Thirty-one studies (n=1208) with 22 from HIC (n=602), reported series of children presenting with MIS-C. Fever, abdominal pain and diarrhoea were the most common symptoms. Nearly half of children (638/1208) who met criteria for MIS-C were admitted to ICU (449/638, 70.3% of which were from HIC) compared with 22.9% in the overall analysis (online supplemental tables 6 and 7).
Disaggregated data were available on 184 neonates with fever; inability to feed/lethargy and dyspnoea were the most commonly reported symptoms. Twenty-one neonates (16.6%, 95% CI 11.2% to 23.9%) were asymptomatic at the time of diagnosis.
Global knowledge of COVID-19 epidemiology, clinical characteristics and management has continued to evolve since the onset of the pandemic. Children have been noted to have relatively lower rates of severe illness and low mortality; however, they have been impacted by MIS-C.4 139
The findings of our review, the largest in terms of published systematic reviews on paediatric COVID-19, are consistent with previous reviews that identified predominance of infection in school-age children, slight male predisposition, prevalence of comorbidities among children with COVID-19 and low hospitalisation and mortality rates.2 4 140 The clinical presentation in children is heterogeneous, including a wide spectrum of clinical features. Fever and cough were the most commonly reported presenting symptoms, in line with the previously published systematic reviews.4 141A U-shaped curve of severity has been demonstrated in children diagnosed with COVID-19 with infants under 1 year of age and adolescents 10–14 years of age at higher risk of developing severe COVID-19.3 47 75 Due to lack of age-disaggregated data, we could not reliably compare the frequency of severe cases by age group in this review. Reported risk factors for severe disease among children include age, viral load142 and presence of comorbidities.3 There is a possibility that children with comorbidities may have been hospitalised related to their underlying chronic condition and incidentally determined to have COVID-19 infection or investigated more extensively. Some of the common comorbidities reported in children with COVID-19 infection include asthma, immunosuppression, congenital heart disease, kidney disease and obesity.3 4 47
Regional differences were identified in the comparison of clinical features, treatment and outcomes between HICs and LMICs. Pooled estimates of hospital mortality were higher in LMICs compared with HICs. Given that it was not possible to calculate risk-adjusted mortality rates for COVID-19, it is unclear whether observed differences in mortality are related to selection bias (eg, differences in severity of illness of included patients or differences in case definitions and inclusion criteria) or differences in available hospital resources. Nevertheless, there is ongoing concern that, in LMICs with high burden of illness and health system limitations, children with severe disease and MIS-C may be at greater risk for adverse outcomes and death than perceived to date. The differences in frequency of observed clinical features may be related to increasing recognition of MIS-C over the course of the pandemic and their inclusion in more recent COVID-19 case series, but is likely similar between HICs and LMICs.
Comparisons of clinical features and outcomes according to severity of illness were limited by heterogeneous reporting across the included case series. A higher proportion of children with severe disease demonstrated symptoms consistent with MIS-C (fever, abdominal symptoms, rash, neurological symptoms, conjunctivitis) and received IVIG, steroids and inotropes.
Compared with previous reviews, several at an earlier stage of the pandemic,4 7 8 140 143 this review has several strengths. Using a broad search strategy implemented in English, Chinese and Spanish databases, we summarise evidence from 129 studies from 31 different countries, the largest sample to-date. We excluded case reports to minimise selective reporting of extreme and atypical cases. We also attempted to reduce possible overlap in cases to prevent duplications. We identified differences in features from studies in HICs compared with LMICs, and between severe and non-severe cases, although with limited available data. Finally, we report subgroup analyses for neonates, and children presenting with MIS-C.
The review is limited primarily by the small sample sizes of individual studies, limitations in study reporting, and study quality limitations due to non-consecutive patient enrolment, unclear case definition and incomplete follow-up to hospital discharge. Our approach of pooling proportions is subject to bias and wide confidence-intervals due to small study sample size. We could not undertake multivariate analysis to identify risk factors for severe infection or adverse outcome in children due to lack of individual-patient-data. The inclusion of asymptomatic cases could have contributed to underestimation of the prevalence of clinical characteristics and optimism in the reporting of outcomes. Finally, it should be noted that a large number (36/129, 28.0%) of the included studies were from China. While the Chinese healthcare system is well-resourced in certain regions, many of the Chinese studies included were conducted in the city of Wuhan or in Hubei Province (n=9, 32.1%), where the gross domestic product per capita is less than half of that of Beijing and Shanghai.144 Therefore, the findings of studies from China may be generalisable to the socioeconomic and health development status of other middle-income countries.
This review contributes to the global understanding of paediatric COVID-19 disease and supports priority setting in research for current pandemic and future outbreaks. This body of literature would be improved by complete reporting of larger series with consecutive recruitment of patients, specific case definitions and complete long-term follow-up to determine global epidemiological trends, age‐specific burden of disease and illness trajectory following COVID-19 infection. Improved characterisation of disease severity and increased reports from low-income countries are needed to better understand differences in clinical manifestations, resource utilisation and outcome by region, which can be integrated in future updated analyses. The concern for selection bias remains as it is possible that in LMICs, the population of hospitalised children was sicker and at higher baseline risk of death, independent of resources. Individual-patient-data meta-analysis would be of benefit to characterise risk factors for severe disease, clinical features in different age groups and account for observed differences in outcome. With respect to clinical management, none of the therapies instituted in the treatment of children with severe COVID-19 disease have been demonstrated to improve outcome in randomised trials; therefore, a recommendation regarding their use is challenging. Given that children appear less likely to develop severe respiratory disease, but are at risk of multiorgan dysfunction due to MIS-C, further studies are needed to characterise the clinical trajectory of this novel syndrome and determine the optimal treatment for it. Finally, there remains paucity of studies reporting long-term prognosis of COVID-19 in children.145
Our review suggests that children predominantly contracted mild form of infection but could be at risk of more severe outcomes. It is crucial to take into consideration risk factors including contact-exposure, underlying comorbidities, young age and male sex which may increase the risk of severe disease. While we have identified several elements that highlight the disease spectrum and higher risk of adverse outcomes in certain settings, such as LMICs, there is the need for much closer scrutiny of this illness globally with individual patient data analysis.
Data availability statement
Data are available upon request