Paediatric COVID-19 mortality: a database analysis of the impact of health resource disparity

Introduction

The clinical presentation, severity and outcomes of acute COVID-19 are different in children compared with adults. While a higher proportion of children are asymptomatic or less severely ill than in many adult reports, severe manifestations do occur.1 While cardiac compromise in the form of multisystem inflammatory syndrome in children (MIS-C) is often described, acute respiratory distress syndrome (ARDS) and other organ dysfunction also occurs in children.2 3 The risk factors for severe disease in in paediatrics are incompletely understood.4 Furthermore, there is a lack of global data to improve understanding of the COVID-19 burden in children who live in low-income to middle-income countries (LMICs) versus those in high-income countries (HICs) sites.5

One systematic review that summarised the difference in paediatric COVID-19 morbidity and mortality in HICs and LMICs has been published.6 However, most studies’ samples analysed fewer than 100 patients, and over half the data came from the USA and China. Furthermore, due to heterogeneous reporting of data in the included studies, the authors were limited in their ability to pool or compare data on clinical features and outcomes. A recent scoping review reported that ethnicity and socioeconomic condition were under-represented in COVID-19 epidemiological studies.7 We aimed to investigate the differences in survival of paediatric patients in LMICs and HICs and the factors that may contribute to such differences between regions.

Methods

Data were collected from hospitalised patients under 19 years of age with confirmed or clinically suspected COVID-19 between 1 January 2020 and 31 March 2021 admitted to institutions across the globe contributing to the International Severe Acute Respiratory and Emerging Infections Consortium (ISARIC) database according to ISARIC/WHO Clinical Characterisation Protocol for Severe Emerging Infection.1 8 Fields for analysis were extracted from the complete dataset gathered from Research Electronic Data Capture (V.8.11.11, Vanderbilt University, Nashville, Tennessee, USA).8

Variables of interest were classified into four domains: comorbidities, presenting signs and symptoms, complications and treatments. Comorbidity was defined as any history of pre-existing medical conditions that were not otherwise related to COVID-19 natural history and reported at admission date. Complications referred to any medical condition detected during patients’ stay that was not present at admission. Therapies included were drugs, oxygen and use of other treatments such as invasive mechanical ventilation (IMV). Ethnicity was collapsed into five categories (black or African American, white, Asian, mixed/others and missing/unknown) following the Centers for Disease Control and Prevention National Health Interview Survey glossary.9 Country income groups were dichotomised into HICs and LMICs according to the latest World Bank classification.10

Date of admission was defined as the date of hospitalisation. Descriptive statistics were described as frequencies (n) and proportions for categorical data, mean±SD or median (IQR) for continuous data, and number of available data (N) for each variable. Demographic characteristics, comorbidities, complications and treatments were compared between country income groups using χ² or Fisher’s exact test as indicated. Kaplan-Meier survival curves were plotted and compared using the log-rank test. Multivariable Cox proportional hazards regression models were fitted to identify mortality predictors. In-hospital survival analysis was performed to obtain 28-day and 90-day survival rates. Intensive care unit (ICU) admission and IMV requirement served as proxies for morbidity. Time-to-event analyses were performed to identify morbidity and mortality defining periods: (1) from hospital admission until ICU admission, (2) ICU admission to first intubation, (3) ICU admission to ICU discharge or death, (4) ICU discharge to hospital discharge or death. A p value of <0.05 was considered statistically significant. Statistical analyses were performed using SPSS V.25 (IBM Corp).

Results

There were 12 860 children, originating from 36 countries, 15 of which (41.7%) were LMICs and 21 (58.3%) HICs(table 1). Seventy per cent (n=9041) of participants were from LMICs and 29.7% (n=3819) cases were from HICs. The majority of participants were contributed from 670 cites in South Africa (59.3%) and 346 sites in the UK (24.1%), followed by Malaysia (4.9%), Poland (2,4%) and Malawi (2.3%) (figure 1 and table 1). COVID-19 status was laboratory confirmed in 69.7%; this rate was higher in HICs (73.0%) in comparison with that in LMICs (68.3%) (table 2).

Request permission

Adolescents aged between 12 and 17 years were the largest age group in LMICs (25.6%) and overall (24.0%), followed by infants younger than 1 year of age (21%). Infants were the largest group in HICs (29.4%). Males (50.8%) and females (49.2%) were similarly represented. Black or African-American participants formed more than one-third of the study population and 52% of LMICs children (table 2).

In total, 425 (3.3%) participants died. The mortality rate was higher in in LMICs compared with HICs (4% vs 1.7%) (table 2). Reported mortality in the UK was 52 children (1.2%), which does not differ significantly (p=0.98) from the rest of HICs. Mortality in South Africa, numbering 268 children (3.5%), differed significantly (p<0.001) from 93 (6.5%) mortality in the rest of LMICs. The total admissions and mortality rates across the study period based on country groups (UK, South Africa and other countries) were presented in figure 2. The curves showed bimodal with peaks of mortality rates coinciding with admissions during July 2020 and January 2021. Available case data in early 2020 was limited compared with later time points. In contrast to South Africa and other countries, mortality rate in the UK did not rise significantly with the first and second wave of case admissions during April–May 2020 and October 2020–January 2021.

Request permission

Children in LMICs had significantly greater and earlier mortality (adjusted HR (aHR) (95% CI) 4.73 (3.16 to 7.10), p<0.001). The 28-day and 90-days survival among all participants were 96.7% (10 339/10 692) and 96.5% (10 642/11 024), respectively. Survival was higher at both 28 (98.3% (3049/3101)) and 90 days (98.1% (3137/3199)) in HICs than LMICs (96.0% (7290/7591) and 95% (7505/7825), respectively).

The availability of information on comorbidities varied between countries of origin. The prevalence of several comorbidities was significantly higher in HICs including chronic neurological disease, seizures, diabetes and chronic cardiac disease. Infectious diseases such as tuberculosis and HIV/AIDS were significantly higher in LMICs (table 3).

The most common presenting symptoms overall include fever (20%), cough (16.1%) and shortness of breath(10.5%). Between-group’s comparisons showed that cough (HIC 14.0% vs LMIC 26.1%), shortness of breath (9.2% vs 16,4%), runny nose (4.2% vs 8.6%), loss of smell or taste (0.7% vs 2.9%), anorexia (0.1% vs 0.8%) and inability to walk (0.1% vs 0.2%) were more common among patients in LMICs.

Complications during hospitalisation are shown in table 4. Patients in LMICs had more bacterial and cryptogenic pneumonia and ARDS, as well as other organ involvement such as brain with stroke or heart with cardiac arrest. The rates of complications were generally higher than that of HICs, except for cardiac arrhythmia and disseminated intravascular coagulation in which HICs rates were statistically higher.

The two most commonly administered therapies were antibiotics in 41.2% of participants, followed by corticosteroids in 11.4%. Antivirus was administered in 5.3% of subjects, with higher percentage observed in HIC, especially of remdesivir, which is more than twice that of LMIC. Adjunctive and supportive treatments were generally performed more often in HICs. No participants in LMICs were treated with extracorporeal membrane oxygenation, as compared with 10 participants in HICs (table 5).

Participants in LMICs were most often admitted to the ICU within the first day of admission, with those who died being admitted earlier than survivors (LMICs vs HICs: 0 (0–1) vs 0 (0–2.5) day respectively, p=0.03) . While time to IMV was not significantly different for survivors versus non-survivors in either LMIC or HICs, most received IMV in the first day of admission and all within a maximum of 3.5 days. Time from ICU admission to ICU discharge or death was significantly shorter in survivors versus non-survivors, although non survivors in HICs had longer stays than in LMICs (3 (1–6) vs 9.5 (4.5–19.2) days respectively in HIC and 0 (0–4) vs 4 (0.7–7.2) days respectively in LMIC, p<0.05). Mortality was significantly higher (p<0.001) in LMICs compared with HICs for participants who received IMV (43.6% vs 7.3%) or who required ICU admission (16.7% vs 3.5). Nearly 2.3% and 6.1% participants died without ICU and/or IMV support in LMICs in comparison with the respective 1.3% and 1.4% in HICs (table 6).

Multivariable analysis was done to evaluate factors associated with mortality. Significant risk factors found were: aged <1 year (aHR (95% CI)=1.80 (1.01 to 3.22)), low-middle income group (4.73 (3.16 to 7.10)), comorbidities such as chronic kidney disease (3.74 (2.20 to 6.35)) or cardiac disease (2.42 (1.50 to 3.91)) and invasive mechanical ventilator requirement (3.46 (2.27 to 5.28)) or exposure to antibiotics (2.07 (1.34 to 3.22)). The use of antiviral agents (aHR=0.55 (0.32 to 0.96)) was the only factor inversely associated with mortality (figure 3).

Request permission

Discussion

We present a large international cohort of children hospitalised with COVID- 19. We found that mortality was significantly higher in LMICs in comparison with HICs. Disparate care patterns were also observed, with patients in LMICs reported to receive most adjunctive and supportive therapies less frequently than patients in HICs. While these findings may represent differences in practice, they may also represent variation in available supports for children based on income status of the country. Such disparities have been described in adult COVID-19 patients, but limited data exist for children. Prior reports have focused on specific aspects of illness such as infection or cardiac dysfunction, have included small cohorts of children or are limited to certain countries or regions.11–20

While the findings may be criticised as mainly representing data from two countries, the UK and South Africa (SA), these countries are good examples of HICs and LMICs. Statistical analysis showed no significant difference of mortality between UK and the rest of HICs. Although mortality in SA was significantly lower than the rest of LMICs, both mortalities from SA and non-SA LMICs were significantly higher than HICs group. Low number of subjects from non-SA LMICs was disproportionate to that of SA, thus conclusion can not be drawn from observed difference in mortality between them. Furthermore, as data supplied from the UK and South Africa comes from national COVID-19 research databases recruiting from a high number of sites in the UK and South Africa, and in this sense may be more representative of country income differences, as opposed to enrolment of single sites (eg, a national referral centre) in different countries. The inclusion of children from many other countries, although relatively small cohorts form each country in comparison, does allow understanding of care patterns in areas around the world.

More participants from HICs could be admitted to ICU and received IMV than LMICs patients. While the small numbers available for analysis in some categories limit our confidence in these findings, in LMICs they do given IMV and dying within shorter periods of time than HICs. Not only were children in LMICs hospitalised with COVID-19 more likely to die, they were also shown to die earlier in their hospitalisation. Despite possible confounding effects from missing data relating to severity of illness at presentation (vital signs, organ failure scores), a positive association between LMICs and mortality were consistently observed in analysis of children admitted to the ICU and those receiving IMV.

Several independent risk factors for mortality were identified in addition to country economic group. Mortality was lowest for patients aged between 1 and 5 years and higher among patients of age <1 or >5 years. This finding confirmed the U-shaped mortality pattern shown in several other reports, although infancy is not always recognised as a risk factor in small studies.1 21 22 Comorbidities such as chronic kidney and cardiac diseases were also shown to be independent risk factors as reported by others.19 23 24 These risk factors were more prevalent in HICs and thus did not seem to be associated with the high mortality rates noted in LMICs, but this may reflect underdiagnosis possibly due to lack of diagnostic resources in some LMICs.20 Higher rates of certain comorbidities, many infectious in nature, is another possible cause of the relationship between LMICs and mortality. Chronic respiratory failure has been associated with death in COVID-19 adult patients, with some evidence that this is occurs in paediatric cases as well. Our data also provided information on tuberculosis, which has not specifically identified as comorbidity in children with COVID-19 in other reports. Similarly, data on the impact of HIV in children is sparse, and our review finds this to be an important risk factor and more prevalent in LMICs.23 25

More patients receiving antiviral therapy were found in HICs versus LMICS. In fact, remdesivir, which is recommended for severe hospitalised COVID-19, was used in exceptionally lower percentage of LMIC subjects. We can only speculate that period of this study occurred when evidence based on antiviral efficacy was still scarce especially in children, or it may indicate lack of access to drug or lack familiarity with recommendations.26 The recommendations for antiviral use in children with severe COVID-19 from the National Institutes of Health have suggested use for patients over the age of 12 years; this recommendation would not have applied to infants who had a higher risk of mortality in our study.27 28 Further investigation of the impact of antivirals in children of all ages should be considered. The efficacy and the cost benefit of these expensive medications in resource-limited sites are needed; if valuable, improving access should then be at the core of discussions.

The use of other therapies also highlight the differences between LMICs and HICs. There was less use of many adjunct therapies associated with outcomes in adult studies such as prone positioning, high flow nasal cannula, antibiotics and steroids in LMIC sites. Whether these differences were the result of lack of availability of therapies or other regional factors cannot be determined, but it seems likely that limitation to access may influence practice. The high rate of mortality in patients outside the ICU and who died without IMV suggests that limitations to ICU beds or ventilators in LMICs likely play a large role in the excess death rates reported as compared with HICs.29 30

The higher prevalence of complications of respiratory disease such as ARDS, bacterial and cryptogenic organising pneumonia, and the impact of organ dysfunction outside the lung such as increased rates of myocarditis, pericarditis, endocarditis, meningitis, encephalitis, stroke and cardiac arrest observed in LMICs are also likely factors in the high death rates. Patients with MIS-C were not specifically reported in the time period of this report.

Our study has the strength of a common reporting format in participating centres around the world. We describe a relatively large number of children and are able to provide both comparisons of patient characteristics and outcomes and evaluate risk factors for outcomes using common definitions. Limitation of this study includes a predominance of patients from one LMICs and one HICs, South Africa and UK, potentially limiting the generalisability of our findings to all countries. In addition, we did not adjust for pandemic era. Inevitably, we have missing data for a number of variables, including comorbidities, which limits the effective sample size of analyses examining relationships with patient characteristics and outcomes. Lack of data on nutritional status of children on each group, which may explain disparity between country income groups, was another limitation of the study. Moreover, considerable proportion of non-confirmed cases also limits the impact of this study on public health policy.

In conclusion, we found many differences in characteristics, treatments and outcomes among children from LMICs and HICs with infants had higher death rates than other children. Patients less frequently receive IMV and other supportive therapies in LMICs, which likely represents disparities in access to healthcare that influence outcomes. Reducing the gap in our ability to care for sick children in LMICs versus HICs will inevitably improve global outcomes during both pandemic and interpandemic periods.