Background
Globally, an estimated 3.3 million RSV-associated lower respiratory tract infection (LRTI) hospitalizations and between 26,300 and 101,400 deaths occur annually in children aged < 5 years. A large proportion of these hospitalizations and deaths occur in low and middle-income countries (LMIC) [
1].
New RSV prevention technologies in the pipeline include maternal vaccination and long-acting monoclonal antibodies (MAB) [
2]. Describing the burden of RSV-associated mild and severe disease is pivotal for estimating the cost-effectiveness of these new interventions. Although incidence data of hospitalization with RSV-associated LRTI from South Africa have been published these data did not include national estimates and did not quantify non-medically attended disease burden nor mild illness associated with RSV [
3]. Burden estimates in fine age groups (in infants < 1 year) including non-medically attended illness will also be used to improve the parameterizing of cost-effectiveness models, which will be valuable for policy makers.
Additional burden of disease may lie in mild illness, and therefore leaving this group out of estimates may significantly underestimate the burden, cost, and cost-effectiveness of interventions [
4]. Due to polymerase chain reaction (PCR) testing the identification of viruses in the respiratory tract has become more sensitive, the presence of these viruses may not be associated with disease, therefore adjusting for the attributable fraction of RSV strengthens burden estimates (paper submitted to the journal of infection). The burden of non-medically attended RSV-associated illness has not been estimated in our setting; however, estimates for influenza suggest that there may be a significant burden of disease in non-medically attended illness [
5]. Data published in our setting, model out of hospital death in the < 5 years age group, refining these estimates to finer age bands will improve mortality burden estimates [
6]. Describing the seasonality of RSV may assist with the implementation of RSV prevention technologies by targeting immunizations and MAB administration prior to and during the peak RSV transmission season [
2].
We aim to describe the full burden of RSV-associated illness in South African children aged < 5 years (both medically and non-medically attended illness) in 1-month age groups for infants and 3-month age groups until 2 years and then yearly until < 5 years; specifically, we describe the burden of RSV-associated acute respiratory illness (ARI), severe acute respiratory infection (SARI) and mortality (in- and out-of-hospital) in South Africa during 2011 to 2016.
Discussion
We were able to document the national burden of RSV-associated illness in young infants by month of life for both mild and severe RSV-associated illness in a LMIC, specifically demonstrating the highest burden of illness is in the first and second month of life. This analysis also describes the significant burden of non-medially attended RSV-associated illness and is the first to document this burden on our setting. This more inclusive description of burden supports interventions such as maternal vaccination and long-acting MABs which target the younger infant. Our analysis adds methodologic techniques that provide more useful data on the burden of RSV in South Africa: estimating medically and non-medically attended illness in each of these narrow age bands, adjusting estimates for RSV cases without fever, and using RSV-AFs to adjust the observed proportion of cases for a more robust estimate.
The burden of disease for mild illness is highest in the 2-month age group and the burden of NMA mild illness is higher than MA mild illness in all age groups. This likely reflects health-seeking behavior and access to primary health care in our setting, where many people may not seek care if illness is perceived as mild even though healthcare is free to children < 5 years. This likely reflects the health-seeking behavior and access to primary health care in our setting, where healthcare is free to children < 5 years. It also highlights the high burden of mild illness in communities and may support an intervention for older children. We adjusted RSV-associated ILI for the proportion of children who present with RSV-associated illness without fever. Nyiro et al. [
18,
20] describe that up to 25% of children with RSV-associated mild illness present without fever. The adjustment assists in describing the substantial burden of RSV-associated in young children. While several estimates of RSV-associated mild or outpatient illness have been published from LMIC, few provide data in fine age bands. In Kenya, a similar methodology to ours Emukule et al. [
4] described the outpatient burden of RSV-associated ARI including an estimate of non-medically attended RSV. The incidence of NMA RSV-associated ILI (6.0/1000 (95% CI 5.4–6.5) in their < 5-year age group was higher compared to 4.7/1000 (95% CI 3.3–66) for medically attended ILI, these estimates are similar to our estimates in this broad age group (< 5 years) but do not allow for comparison in finer age band [
4]. This higher proportion of MA illness in our setting, may be due to cost-free access to care for children.
The burden of RSV-associated severe illness is highest in the first month and second month of life after which the burden declines to 11 months of life and further through years 2 to 4 of life. Young age is a well-described risk factor for RSV-associated severe illness and infants < 6 months will be the group of infants that will most benefit from prevention interventions (maternal vaccine of MAB) [
22]. Although the burden of NMA severe illness is higher than MA severe illness there is still a significant burden of illness in this group. The burden of NMA attended severe illness may be due to socioeconomic factors impeding access to care. Similar to the severe RSV-associated illness burden, mortality due to RSV-associated respiratory illness is highest in the first 3 months of life, declining steadily until 11 months of life, with few deaths in the older age groups. There are data on the burden of RSV-associated severe illness in other LMICs, but most of these are in wider age bands than what we present here. Li et al. [
1] published a meta-analysis of global RSV-associated hospitalization incidence; these results reflect a large burden of disease in infants less than 4 months of age in LMIC countries (31/1000, 95% CI 17.0–56.4) and are similar to our estimates in the same age group. However, our data estimates the burden in < 1-month, < 2-month, and < 3-month age groups, illustrating how larger age bands dilute the substantial burden in the first two months of life. Our estimates are supported by an analysis done in South Africa which estimates the hospitalization rate for RSV-associated illness to be 7910/100,000 population (95% CI 6155–9665), very similar to our MA RSV-associated severe illness estimates of 6804/100,000 (95% CI 5065–8529) in the < 1 month of age and through the first year of life [
3]. A cohort study from South Africa, reports incidence rates in the first 2 years of life (RSV-associated illness) of 0.07 cases per child year (95% CI 0.05–0.10) for hospitalized RSV-associated LRTI in children < 6 months. These estimates are very similar to the estimates in this study. With these comparable estimates in hand, we provide policy makers with evidence to consider RSV-prevention interventions to protect young infants in our setting [
23]. In Kenya, using a similar methodology to ours, the burden of disease for severe cases included NMA severe RSV-associated illness in children aged < 1 year was 14.5/1000 (95% CI 8.9–23.7) which is lower than our estimates of 3752/100,000 (95% CI 2530–5357) [
4]. However, our estimates for non-medically attended mild illness (6657/100,000 (95% CI 2692–12,540)) were higher than the estimates for mild non-medically attended illness in Kenya 8.9/1000 (95% CI 4.8–16.7) [
4]. The difference may be due to different health-seeking behaviors in the different settings.
An additional strength of our study was the ability to adjust numbers by the AF-RSV, described in our setting [
17]. This is important because the detection of respiratory viruses does not always reflect their role in causing disease [
17,
24]. While the AF for RSV was generally high in infants, in children aged 1–4 years the AF-RSV was 74.6% (95% CI 53.6–86.0%) for mild illness and 83.4% (95% CI 70.9–90.5%) for severe illness. This adjustment refined our estimates in these age groups.
Although the case fatality ratio (CFR) for RSV-associated illness in infants (CFR between 1.8% (95% confidence interval (CI) 0.8–3.6%) and 1.0 (0.4–1.5) infants < 3 months and those 3–6 months respectively) is lower than other viral pathogens, such as influenza (CFR 3.2% (0.6–15.4%) [
1,
25,
26]. Mortality in the young infants will increase years of life lost (YLL) and affect cost-effectiveness estimates of intervention to prevent RSV-associated severe illness. Out-of-hospital mortality is difficult to define, specifically where the cause of death and death registration is not documented. In an earlier analysis of excess mortality attributable to RSV, we estimated that approximately 26% of RSV-associated deaths in children aged < 5 years occur in the community [
21].That model estimated 665 (95% CI 105–1105) deaths in children aged < 5 years to be RSV-associated very similar to this estimate of 650 (95% CI 479–947) deaths in children < 5 years. Shi et al. estimated that up to 49% of deaths from RSV-associated SARI, in children < 5 years, in LMIC countries occur out of hospital [
27]. Other estimates from recent publications suggest that out-of-hospital deaths may account for a larger proportion of RSV deaths for example in India > 80% of RSV-associated deaths occur in the community, in Zambia the estimate was 62% and in Pakistan 27% [
28‐
30]. There are many differences between these settings, including urban vs rural settings, cost of accessing health care, and healthcare system factors. Nonetheless, these newly published data suggest that our estimate may be a minimum estimate.
Variation in RSV seasonality by country, region, and climatic zone is well described, implying that the description of seasonality in each country is important [
31,
32]. Describing country-specific seasonality will assist with decision-making regarding the seasonal vs all-year administration of interventions such as maternal vaccine and MAB treatment for young infants. Knowledge of RSV-seasonality will also provide important data for cost-effectiveness models.
Limitations of our study include: This method is sensitive to the description of the mild and severe cases in the base catchment area. Even though we made adjustment for non-enrolment, conducted a HUS and our base hospitals had defined catchment populations it is possible that cases were missed. While we were able to make use of observed data in many provinces (5/9), not all provinces of the country were included in the observed data, this may not account for the difference in the detection rate of RSV between provinces. The difference in the prevalence of risk factors from ecological data such as the DHS and published data may not account for all the differences between provinces. In the same way, health-seeking behavior may differ between rural and urban sites leading to over or underestimation of NMA mild and severe illness based on a HUS done at 3 sites. Population data were adjusted up from the 2011 census and may not accurately account for population movement and increases. Our analysis produced a wide confidence interval in some age groups, this is likely due to the low numbers in these finer age bands and the adjustment used from observed cases.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.