The national burden of influenza‐like illness and severe respiratory illness overall and associated with nine respiratory viruses in South Africa, 2013–2015

Abstract Background Estimates of the disease burden associated with different respiratory viruses are severely limited in low‐ and middle‐income countries, especially in Africa. Methods We estimated age‐specific numbers and rates of medically and non‐medically attended influenza‐like illness (ILI) and severe respiratory illness (SRI) that were associated with influenza, respiratory syncytial virus (RSV), rhinovirus, human metapneumovirus, adenovirus, enterovirus and parainfluenza virus types 1–3 after adjusting for the attributable fraction (AF) of virus detection to illness in South Africa during 2013–2015. The base rates were estimated from five surveillance sites and extrapolated nationally. Results The mean annual rates per 100,000 population were 51,383 and 4196 for ILI and SRI, respectively. Of these, 26% (for ILI) and 46% (for SRI) were medically attended. Among outpatients with ILI, rhinovirus had the highest AF‐adjusted rate (7221), followed by influenza (6443) and adenovirus (1364); whereas, among inpatients with SRI, rhinovirus had the highest AF‐adjusted rate (400), followed by RSV (247) and influenza (130). Rhinovirus (9424) and RSV (2026) had the highest AF‐adjusted rates among children aged <5 years with ILI or SRI, respectively, whereas rhinovirus (757) and influenza (306) had the highest AF‐adjusted rates among individuals aged ≥65 years with ILI or SRI, respectively. Conclusions There was a substantial burden of ILI and SRI in South Africa during 2013–2015. Rhinovirus and influenza had a prominent disease burden among patients with ILI. RSV and influenza were the most prominent causes of SRI in children and the elderly, respectively.

Conclusions: There was a substantial burden of ILI and SRI in South Africa during 2013-2015. Rhinovirus and influenza had a prominent disease burden among patients with ILI. RSV and influenza were the most prominent causes of SRI in children and the elderly, respectively.

| INTRODUCTION
Despite the use of interventions such as antimicrobial drugs and vaccination against leading pneumonia-causing pathogens, acute respiratory infections (ARIs) remain a major cause of death globally, especially among children aged <5 years. 1 In addition, mild and severe-non-fatal episodes of ARI are responsible for a substantial burden on the healthcare systems and the society, through illness, absenteeism, and associated costs.
The Pneumonia Etiology Research for Child Health (PERCH) study conducted in seven low-and middle-income countries, including South Africa, estimated that respiratory viruses were responsible for 61% of severe pneumonia cases (compared with 27% due to bacteria) among HIV-uninfected children aged <5 years. 2 This suggests that, as bacterial etiologies decline due to vaccination, respiratory viral causes may gain greater prominence. 2 Nonetheless, the relative contribution of respiratory viruses in individuals of all ages remains poorly understood.
HIV infection is associated with increased severity of ARI and higher case-fatality ratios, especially in older children and younger adults. 3,4 In South Africa, a country with an HIV prevalence of 12.7% in the general population and 23.5% in individuals aged 25-44 years in 2016, 5 pneumonia and influenza were the leading causes of death among children aged <5 years and the 3rd and 5th causes of death among older individuals, resulting in 19,638 deaths annually in the same year. 6 Whereas the etiology of hospitalized severe pneumonia among HIV-uninfected children aged <5 years has been described in South Africa, 2 the overall burden on the healthcare system and society of mild and severe respiratory illness overall and associated with respiratory viruses across age groups has not been quantified. Quantifying this burden may assist policy makers with allocation of resources and prioritization of interventions.
In this study, we sought to assess the mean annual national burden of medically and non-medically attended influenza-like illness (ILI) and severe respiratory illness (SRI) overall and associated with nine respiratory viruses in different age groups in South Africa during 2013-2015. with symptoms of any duration who met age-specific clinical inclusion criteria. A case in children aged 2 days to <3 months included any hospitalized patient with diagnosis of suspected sepsis or physiciandiagnosed acute lower respiratory tract infection irrespective of signs and symptoms. A case in children aged 3 months to <5 years included any hospitalized patient with physician-diagnosed acute lower respiratory tract infection, including bronchitis, bronchiolitis, pneumonia, and pleural effusion. A case in individuals aged ≥5 years included any hospitalized patient presenting with manifestation of acute lower respiratory tract infection with fever (≥38 C) or history of fever and cough.
In addition, we conducted prospective surveillance for patients presenting with ILI at two outpatient clinics (Edendale Gateway Clinic, KwaZulu-Natal province and Jouberton Clinic, North West province) located in the same catchment area as the abovementioned hospitals over the same study period. An ILI case was defined as an outpatient of any age presenting with either temperature ≥38 C or history of fever and cough of duration of ≤10 days.
We also enrolled persons presenting at the same outpatient clinics with no history of fever, respiratory, or gastrointestinal symptoms during the 14 days preceding the visit (hereafter referred to as controls). These individuals commonly presented to the clinics for visits such as dental procedures, family planning, well baby visits, voluntary HIV counseling and testing, or acute care for non-febrile illnesses. We aimed to enroll one HIV-infected and one HIV-uninfected control every week in each clinic within each of the

Data source 4: Population denominators
We obtained age-and year-specific population denominators for the catchment area of the surveillance sites described in Data source 1 from projections of 2011 census data for South Africa. 15 We also obtained the provincial age-and year-specific population denominators from the same data source.

| Estimation of the AF of respiratory viruses detection to illness
We used unconditional logistic regression to estimate the attributable fraction (AF) of respiratory viruses-associated hospitalization and outpatient consultation by comparing the respiratory viruses detection rate among ILI or SRI cases with those of controls (outcome variables in the models). The AF was estimated from the virus specific odds ratio (OR) obtained from the multivariable regression models as follows: Subsequently, we estimated the individual respiratory viruses detection rate associated with illness among ILI and SRI cases (Infl DetectRateIll ) from the observed detection rate (Infl DetectRateObs ) as follows: This analysis was implemented overall and within the following age categories: <5, 5-44, and ≥45 years of age. All estimates were obtained from multivariable models that included the different respiratory viruses investigated in this study, HIV infection, and underlying medical conditions as covariates as previously described. 16

| Medically attended illness
To estimate the national number of respiratory viruses-associated SRI hospitalizations, we used a four-step approach. In Step 1, we estimated the SRI hospitalizations rates at the two hospitals mentioned above (Data source 1) during 2013-2015 as previously described, 18 and we used the SRI hospitalization rates at the two sites as proxy for the corresponding provinces (considered to be the base provinces in our estimation approach). In Step 2, we estimated the SRI hospitalizations rates for the other seven provinces from the base provinces using a previously described methodology that leverages provincial differences in the prevalence of known risk factors for pneumonia and healthcare seeking behavior (Data source 3). In Step 3, we estimated the respiratory viruses-associated SRI hospitalizations rates using available virological surveillance data (Data source 1) (i.e., detection rate of individual viruses adjusted for the estimated AF for each virus). In Step 4, we obtained the number of respiratory viruses-associated SRI hospitalizations using the estimated respiratory viruses-associated rates and the population at risk in each province (Data source 4).
To estimate the national number of respiratory viruses-associated ILI outpatient consultations, we used an approach similar to that used for SRI cases, but we did not adjust for the provincial level risk factors for pneumonia. 18

| Non-medically attended illness
To estimate the national number and rates of respiratory virusesassociated non-medically attended ILI and SRI, we used the four-step approach described above in conjunction with HUS data (Data source 2). 18

| Estimation of the confidence intervals
We obtained the 95% confidence intervals (CIs) for medically and non-medically attended respiratory viruses-associated illness using 3 | RESULTS   (Table 2 and S1).

| Detection of respiratory viruses
The circulation patterns of the respiratory viruses investigated in this study are provided in Figure 1. 13,381 per 100,000 population) were medically attended (Table S3).

| Respiratory viruses AF and AF-adjusted prevalence
The ILI rates were highest in infants and declined with increasing age ( Figure 3 and Table S2). Overall, among patients with ILI, rhinovirus     (Figure 2 and Table S2). A similar ranking was observed among individuals aged ≥5 years ( Figure 2 and Table S2).
Among children aged <5 years, rhinovirus had the highest rate (9424 per 100,000 population) followed by influenza (7440 per 100,000 population) and RSV (4774 per 100,000 population). The respiratory viruses-associated ILI rates were highest in infants and declined with increasing age with the exception of influenza and adenovirus where the highest rates were among individuals aged 5-24 years ( Figure 3 and Table S2) and PIV Type 2 where the highest rate was among individuals aged ≥65 years ( Figure 5 and Table S8). The rates of medically and non-medically attended respiratory viruses-associated ILI are provided in Tables S3, S4, and S8. 1945 per 100,000 population) were medically attended (Table S6).

| Burden of SRI and that associated with respiratory viruses
The SRI rates were highest in infants aged <1 year followed by individuals aged ≥65 years ( Figure 4 and Table S5). Overall, among patients with SRI, rhinovirus had the highest rate (400 per 100,000 population) followed by RSV (247 per 100,000 population) and influenza (130 per 100,000 population) ( Figure 2 and Table S5).
Among children aged <5 years, RSV had the highest rate (2026 per 100,000 population) followed by rhinovirus (1638 per 100,000 population) and adenovirus (667 per 100,000 population) ( Figure 2 and Table S5). Among individuals aged ≥5 years, rhinovirus had the highest rate (253 per 100,000 population) followed by influenza (98 per 100,000 population) and PIV Types 1-3 (39 per 100,000 population) ( Figure 2 and Table S5). The respiratory virusesassociated SRI rates were highest in infants aged <1 year followed by children aged 1-4 years with the exception of influenza where the second highest rate was among individuals aged ≥65 years ( Figure 4 and Table S5) and PIV Type 2 where the highest rate was among individuals aged ≥65 years followed by infants aged <1 year ( Figure 5 and Table S8). The rates of medically and non-medically attended respiratory viruses-associated SRI are provided in Tables S6, S7, and S9.

| DISCUSSION
We estimated the AF of nine respiratory viruses among patients with ILI and SRI as well as the mean annual number and rates of these syn- Zambia (843 per 100,000 population). 23 It should be noted that the SARI case definition used in these studies included patients with a symptom duration of 7 (mostly) or 10 days, whereas our SRI case definition included patients with symptoms of any duration, which would include more patients. In a study conducted in South Africa, the rates of influenza-associated SRI (67 per 100,000 population) were 2.6 times higher than those of influenza associated SARI with symptoms duration of 7 days. 24 Besides the difference in case definitions, difference in rates in different settings can also be attributed to differences In our study, among patients with ILI, the most commonly detected virus was rhinovirus, followed by influenza and adenovirus.
These viruses were detected at high frequencies also in other studies conducted in Africa among outpatients with respiratory illness. [25][26][27][28][29] Rhinovirus, RSV, and adenovirus were the most commonly detected viruses among patients with SRI in our study, and, although variability was observed in different settings, these viruses were frequently detected among hospitalized patients with SARI/ARI in other studies from African countries. [30][31][32][33] Among patient with ILI the AF of influenza, RSV, HMPV, and PIV  This was reported also in another study conducted in South Africa. 10 Rhinovirus has been well described as one of the most common causes of common cold 34 ; however, its role as causal agent of SRI is less well described. In our study, only RSV had an AF-adjusted prevalence higher than that of rhinovirus among children aged <5 years with SRI. that is not associated with illness using controls. When the etiological fraction of different pathogens to a syndrome is estimated, one pathogen is then "assigned as the cause" of the observed illness. Nonetheless, there is a potential inherent "bias" in such approach, in that the observed illness may be the result of the interaction of multiple pathogens. 2

ACKNOWLEDGMENTS
We thank all members involved in SRI and ILI surveillance for the collection of specimens, the data management, and laboratory teams at

DISCLAIMER
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the US Centers for Disease Control and Prevention, USA or the National Institute for Communicable Diseases, South Africa.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1111/irv.12949.