Epidemiology of SARS‐CoV‐2 infection and SARS‐CoV‐2 positive hospital admissions among children in South Africa

Abstract Introduction We describe epidemiology and outcomes of confirmed SARS‐CoV‐2 infection and positive admissions among children <18 years in South Africa, an upper‐middle income setting with high inequality. Methods Laboratory and hospital COVID‐19 surveillance data, 28 January ‐ 19 September 2020 was used. Testing rates were calculated as number of tested for SARS‐CoV‐2 divided by population at risk; test positivity rates were calculated as positive tests divided by total number of tests. In‐hospital case fatality ratio (CFR) was calculated based on hospitalized positive admissions with outcome data who died in‐hospital and whose death was judged SARS‐CoV‐2 related by attending physician. Findings 315 570 children aged <18 years were tested for SARS‐CoV‐2; representing 8.9% of all 3 548 738 tests and 1.6% of all children in the country. Of children tested, 46 137 (14.6%) were positive. Children made up 2.9% (n = 2007) of all SARS‐CoV‐2 positive admissions to sentinel hospitals. Among children, 47 died (2.6% case‐fatality). In‐hospital deaths were associated with male sex [adjusted odds ratio (aOR) 2.18 (95% confidence intervals [CI] 1.08–4.40)] vs female; age <1 year [aOR 4.11 (95% CI 1.08–15.54)], age 10–14 years [aOR 4.20 (95% CI1.07–16.44)], age 15–17 years [aOR 4.86 (95% 1.28–18.51)] vs age 1–4 years; admission to a public hospital [aOR 5.07(95% 2.01–12.76)] vs private hospital and ≥1 underlying conditions [aOR 12.09 (95% CI 4.19–34.89)] vs none. Conclusions Children with underlying conditions were at greater risk of severe SARS‐CoV‐2 outcomes. Children > 10 years, those in certain provinces and those with underlying conditions should be considered for increased testing and vaccination.

died in-hospital and whose death was judged SARS-CoV-2 related by attending physician. Conclusions: Children with underlying conditions were at greater risk of severe SARS-CoV-2 outcomes. Children > 10 years, those in certain provinces and those with underlying conditions should be considered for increased testing and vaccination. of COVID-19 cases globally. [2][3][4] However, the proportion of cases in children varied by country and was as high as 11% in the United States, 23% in Paraguay, and 15% in Brazil. 5 Several factors might contribute to the lower reported incidence of COVID-19 among children: SARS-CoV-2 infection is more likely to be asymptomatic or cause milder symptoms in children (80%) than in adults (40-60%), and children may be less likely to receive medical care or be tested. 2,6-10 Furthermore, some data suggest children under 10 years of age might be 48% less susceptible to infection following exposure compared to adults. 10 Therefore, the testing rates could be lower in that population. 11,12 Published data from the United States suggest rates of admission to hospital and intensive care units (ICUs) have been lower among children with COVID-19 (cumulative hospitalization rates eight cases per 100,000 children <18 years) than adults (164.5 cases per 100,000 adults). 13,14 Systematic reviews of studies published up to May 2020 reported pediatric case fatality ratio (CFR) ranging between 0% (in two studies) and 0.2% across their respective included studies. [15][16][17][18][19] In a study of 114,000 SARS-CoV-2 positive deaths in the United States, children aged <18 years made up <0.3% of all deaths. 20 Subsequent studies of children hospitalized with COVID-19 have described greater risk of severe disease especially among children with obesity and other underlying medical conditions compared to those without these conditions. 13,21,22 There is a paucity of studies describing the epidemiology and clinical features of COVID-19 among children in Africa. Of published studies, a relatively small number of children were studied, ranging from 34-1,439. [23][24][25] Findings from China, 7 Europe, 15 and North America, 19 where the majority of studies of COVID-19 in children were conducted, may not be generalizable to countries such as South Africa where conditions such as malnutrition, childhood obesity, tuberculosis, HIV infection, or HIV exposure among children are more prevalent and background child mortality rates from other bacterial and viral infections are higher than other countries. 26,27 Additionally, sub-optimal sanitation, overcrowding, and limited access to health care likely reduce the efficacy of non-pharmacological interventions against COVID-19.
We describe the epidemiology of SARS-CoV-2 infection and hospitalization among children aged <18 years in South Africa. Specifically, we describe age-specific population level testing rates, incidence of laboratory-confirmed SARS-CoV-2 infection, and the clinical characteristics and outcomes of children admitted with COVID-19 at sentinel hospitals. We also identify modifiable factors associated with inhospital deaths to guide interventions for this population.

| Setting
South Africa is an upper middle-income country with high income inequality-GINI coefficient of 0.65 at national level. 28 The country had an estimated 2020 mid-year population of 59.6 million, of which 33.5% (an estimated 20 million) were children <18 years old. 29 In 2019, the country's under five mortality rate was 34 per 1,000 live births compared to 2 per 1,000 in the Nordic countries, 6 per 1,000 in the United States, and 117 per 1000 in Nigeria and Somalia. 29,30 The country is divided administratively into nine provinces in which there are wide variations in income and healthcare access and quality. The majority of the population lives in low-income settings characterized by high levels of unemployment and limited access to medical insurance and medical care. South Africa started real time reverse transcription polymerase chain reaction (rRT-PCR) testing for SARS-CoV-2 infection on January 28, 2020, and the first case of SARS-CoV-2 infection was reported on March 5, 2020. As part of non-pharmaceutical interventions to curb the spread of the epidemic, schools were closed on March 18, 2020, and measures restricting non-essential travel and trade were introduced on March 27, 2020.
Restrictions were gradually eased starting May 1, 2020, paradoxically as the epidemic started to surge, with phased reopening of schools from June 8, 2020, with all children returning to school by September 1, 2020.

| Data sources and collection procedures
SARS-CoV-2 rRT-PCR results were reported by both public and private laboratories to a surveillance system coordinated by the National Institute for Communicable Diseases (NICD). Limited demographic and epidemiological data such as age, sex, and contact information were obtained at the time of specimen collection. People meeting the South African National Department of Health (NDOH) case definition for persons under investigation (PUI) were tested. In March 2020, a PUI was defined as a person, regardless of age, with acute onset of fever >38.5 C with one or more of the following: cough, fever, or sore throat and contact with a known case of  This definition was revised several times over the reporting period. For example on June 1, 2020, the guidance around which individuals could be tested by SARS-CoV-2 rRT-PCR was changed to restrict testing to those with symptoms, those who needed admission, and those with underlying conditions. 31 For hospitalized people, data were collected at admission, during hospitalization, and at discharge using the DATCOV system, a prospective surveillance program for sentinel hospitals.  33 Because not all hospitals started reporting admissions at the same time, hospitals which started reporting later in the surveillance period could submit their data retrospectively to include all admissions regardless of age. 32

| Laboratory procedures
Testing for SARS-CoV-2 using rRT-PCR began on January 28, 2020, at the reference laboratory at NICD and was expanded to a national Patients received their results through a short-text messaging system (SMS) directly from the laboratory or through the ordering physician.

| Data management
Data on SARS-CoV-2 rRT-PCR tests conducted and case notifications were extracted from laboratory information systems, while data on hospitalizations were extracted from the DATCOV platform. Once

| Data analysis
Descriptive statistics were used to determine the age-specific testing, percent positive and incidence rates, and case fatality ratios (CFRs) comparing children (age <18 years) to adults (age ≥18 years) overall.
Among children, descriptive statistics were used to describe testing rate, percent positive proportion, incidence rate, and case fatality ratios by age categories, sex, epidemiology week, and location (province). The main outcomes analyzed were (1) the SARS-CoV-2 rRT-PCR testing rate and percent positive, (2) incidence of laboratoryconfirmed SARS-CoV-2 infection, and (3)

| Ethical considerations
The NICD has ethical clearance for essential communicable diseases surveillance and outbreak response investigation activities from the University of the Witwatersrand's Human Research Ethics Committee (Medical) (M160667). This activity was reviewed by CDC and was conducted consistent with applicable federal law and CDC policy.* All personal and identifying information were removed, and case records were uniquely assigned to a system-generated record identifier.

| In-hospital case fatality ratio
Among the SARS-CoV-2 rRT-PCR positive childhood admissions (n = 2,007), the median length of hospital stay was 3 days (IQR 1-7 days). One hundred and fifty-four children (7.7%) were admitted into an ICU, of whom 57 were ventilated (2.8%).  (Table 4). Similar findings were observed using the same multivariable model on unimputed data, although the model run with unimputed data had a lower odds ratio for the association between mortality and having one or more underlying conditions (Tables S3 and S4). Ethnicity was not associated with mortality in univariable analyses and was therefore not included in multivariable models with imputed or unimputed data.

| DISCUSSION
In this study of pediatric COVID-19 in South Africa, we found an overall 14.6% of children (0-17 years) tested using SARS-CoV-2 rRT- Denominator is those with data on underlying conditions. b Only these were included in the analysis for in-hospitals analysis as they had outcome data available.  Iranian studies of hospitalized children. 41,42 The childhood in-hospital mortality in our study was also lower than the 11% reported in a study of hospitalized children from the Democratic Republic of Congo. 23 This finding was notable because the children in our study had a similar age and sex distribution and lower prevalence of underlying medical conditions compared with other hospital studies (16% compared to 40-66% in other hospital studies). 13,21,22 Consistent with other studies, 13,22,43 higher in-hospital CFR was associated with age <1 year.
The CFR in those <1 year was four times higher compared to children For this reason, we were unable to adequately describe the clinical presentation among cases. Additionally, we likely underestimated the total burden of underlying conditions among admitted children with the differential effects overestimated or underestimated. Third, we were unable to review medical records to ascertain and fully classify causes of death among childhood SARS-CoV-2-positive admissions.
As we did not have clinical data or post-mortem data, we were unable Individuals who were SARS-CoV-2 rRT-PCR positive results AND were admitted to sentinel hospital AND had with complete outcome information, that is, either died or were discharged from hospital. b From univariable and multivariable models on imputed data. Multivariable model adjusted for age, birth sex, admission to public or private sectors, and presence of underlying conditions. c Included heart disease, diabetes, malignancy, renal disease, and obesity. In another model adding hypertension, chronic respiratory disease or asthma to definition of non-communicable underlying medical condition; in addition to initial list (i.e. heart disease, diabetes, malignancy, renal disease and obesity), ≥1 non-communicable underlying condition was still associated with mortality risk albeit with lower strength of association: aOR 3.29 (95% CI 1. 19-8.37).
to determine whether the SARS-CoV-2 infection was co-incidental or involved in the casual pathway to death. Fourth, the data on SARS-CoV-2-positive admissions were collected from sentinel hospitals as opposed to all hospitals admitting individuals with SARS-CoV-2. We were therefore unable to calculate population level hospitalization rates. Fifth, we considered the first occurrence of hospitalization and may have missed deaths occurring on re-admissions. Sixth, because our data collection forms were not designed to collect data on the multisystem inflammatory syndrome in children (MISC), we were unable to describe MISC which could contribute towards SARS-CoV-2 morbidity and mortality among children. 46