Age‐specific incidence rates and risk factors for respiratory syncytial virus‐associated lower respiratory tract illness in cohort children under 5 years old in the Philippines

Background Respiratory syncytial virus (RSV) is one of the main viral causes of lower respiratory tract illness (LRTI), especially in young children. RSV vaccines, including maternal and infant vaccines, are under development; however, more epidemiological studies are needed to develop effective vaccination strategies. Objectives To estimate detailed age‐specific incidence rates and severity of RSV‐associated LRTI (RSV‐LRTI) using data from a community‐based prospective cohort study in the Philippines. Patients/Methods Cohort children who visited health facilities due to acute respiratory symptoms were identified, and nasopharyngeal swabs were collected to detect RSV. The severity of RSV‐LRTI was assessed using the severity definition proposed by the World Health Organization. Risk factors for developing RSV‐LRTI and contribution of SpO2 measurement were also evaluated. Results A total of 395 RSV episodes which occurred in children aged 2‐59 months were categorised as 183 RSV‐LRTI, 72 as severe RSV‐LRTI and 29 as very severe RSV‐LRTI. Children aged 3‐5 months had the highest incidence rate of RSV‐LRTI, at 207.4 per 1000 child‐years (95% CI: 149.0‐279.5). Younger age group, place of living and low educational level of caregivers were associated with developing RSV‐LRTI. Clinical manifestations had low levels of agreement with hypoxaemia as measured by pulse oximeter. Conclusion The highest burden of RSV was observed in young infants aged 3‐5 months, whereas the burden was also high in those aged 12‐20 months. Future vaccination strategies should consider the protection of older children, especially those aged one year, as well as young infants.


| INTRODUC TI ON
Human respiratory syncytial virus (RSV), recently renamed as Human orthopneumovirus belonging to the Pneumoviridae family Orthopneumovirus genus, 1 is one of the most important viral pathogens which causes lower respiratory tract illness (LRTI) including bronchiolitis and pneumonia, especially in infants and young children. 2,3 Respiratory syncytial virus is divided into two subgroups, A and B, based on antigenic and sequence variations. The predominant subgroup and proportion of each subgroup vary between epidemics. 4 A recent global estimate showed that the incidence of RSV-associated acute lower respiratory infection was 33.1 million, with 3.2 million hospital admissions and 59 600 in-hospital deaths in children younger than 5 years old in 2015. 5 Despite the burden of this disease, there are no commercially available vaccines and the monoclonal antibody is only a specific measure taken to prevent severe disease in high-risk infants. 6,7 The incidence rate (IR) of RSV-associated lower respiratory illness (RSV-LRTI) is higher in the first 6 months of life compared with older age groups, and the illness is generally more severe in young infants. 5,8 For these reasons, it is believed that natural maternal immunity is not enough to prevent severe RSV-LRTI in early infancy. Different strategies for vaccination, including maternal vaccination and infant vaccination, have been proposed to prevent severe infection. 9,10 However, effective implementation of these strategies depends on understanding the epidemiological patterns of severe infections, including age-specific incidences.
Respiratory syncytial virus infections are associated with various levels of respiratory illnesses, from cold-like mild illness to very severe and even life-threatening LRTI. To evaluate the disease burden and effectiveness of vaccines or other interventions, it is necessary to define the severity of RSV infection. 11 To identify severe LRTI with hypoxaemia, percutaneous arterial oxygen saturation (SpO 2 ) measurement by a pulse oximeter has been shown to be effective. Studies have reported that the recognition of hypoxaemia using a pulse oximeter improved the management of children with LRTI. 12,13 It was also shown that one of the most important indicators for severe RSV-LRTI was an SpO 2 level of less than 90%. 14 Recently, the World Health Organization (WHO) expert group proposed that the measurement of SpO 2 should be included in the severity definition for RSV-LRTI. 11 Several prospective cohort studies have been conducted in low-and middle-income countries to define age-specific IRs. [15][16][17][18][19][20][21][22] However, to date, few studies have reported the IR of RSV-LRTI using standard definitions of severity. Hence, in this study, we analysed the data from a community-based prospective cohort study of children younger than five years old to understand in detail the age-specific incidence of RSV-LRTI, using the definitions proposed by the WHO expert group. 11 We examined possible risk factors for developing RSV-LRTI and severe RSV-LRTI and compared clinical manifestations between severe and non-severe cases using data from a community-based prospective cohort study in the Philippines. Further, we evaluated the level of concordance between the severity of the LRTI definitions and the classification based on the integrated management of childhood illness (IMCI). 23

| Study design
In the previously described prospective cohort study, 24,25 a cohort of 4012 children younger than 5 years old was followed in two municipalities, Kawayan

| Definition of RSV-associated RTI
An RTI was defined as a respiratory episode with cough and/or difficult breathing. A new episode was identified if the onset of the new RTI episode was at least 7 days from the previous episode. An episode of RSV-associated RTI (RSV-RTI) was defined based on positive laboratory results for RSV. If there were two RSV-positive samples within a month, the episode with the latter positive sample was considered as prolonged shedding, except if the RSV subgroups or the partial G gene sequences were different from each other.

| Classification of RTI severity
We categorised RTIs into LRTI, severe LRTI and very severe LRTI based on the severity definitions proposed by the WHO expert K E Y W O R D S cohort study, lower respiratory tract illness, pulse oximetry, respiratory syncytial virus group (Supporting Information Table S1). 11,26 Briefly, LRTI was defined as RTI with fast breathing (≥50 breaths/min in children aged 2-11 months, ≥40 breaths/min in children aged 12-59 months) or SpO 2 < 95%; severe LRTI as LRTI with chest indrawing or SpO 2 < 93% and very severe LRTI as LRTI with inability to feed, sleeping most of the time, difficult to wake or SpO 2 < 90%. When RTI occurred in children aged 0-1 month, the severity could not be defined based on the WHO definitions mentioned above. 11,26 Therefore, for this age group, we categorised RTI with ≥60 breaths/min or SpO 2 < 95% as LRTI and RTI with chest indrawing or SpO 2 < 93% as severe LRTI. We also classified RSV-associated respiratory disease using Integrated Management of Child Illness (IMCI), both with and without the revision proposed in 2014. 26 In this revision, children with pneumonia having chest indrawing without danger signs were classified as "pneumonia" (non-severe disease). Hospitalisation of the children with severe respiratory disease was indicated by local treating physician.

| Viral detection and defining RSV subgroups
Nasopharyngeal swabs were collected from RTI for RSV testing as described previously, 27 including real-time polymerase chain reaction (PCR) to screen for RSV, 28,29 conventional PCR for subgrouping and sequencing of the second variable region of the G gene (RSV-A: 342 bp and RSV-B: 330 bp). 30,31 Testing for RSV and other viruses, including rhinovirus, enterovirus, human metapneumovirus, human adenovirus, parainfluenza virus and influenza virus, was conducted using primers described in Supporting Information Table S2. Details of the procedure were described elsewhere. 24 We did not exclude RSV-RTI with other viruses from the analysis.

| Statistical methods
Proportion of RSV subgroups by severe levels was compared using chi-squared test. Agreement of severity classification between IMCI 26 and the RSV-LRTI severe case definition and agreement of SpO 2 < 93% and having other clinical manifestations were measured using Cohen's kappa statistic. Age-specific IRs were calculated for every three-month age group, except for children aged <3 months.
These IRs were calculated separately for 0-1 month (as undefined RSV-RTI) and 2 months old. The IRs of each age group were calculated as the sum of new episodes divided by the sum of the childyears obtained for each age group. Poisson regression model was used to calculate 95% confidential interval (CI) of IRs. To evaluate the risk factors associated with developing RSV-LRTI, we calculated hazard ratios (HRs) for total RSV-LRTI (including RSV-LRTI, severe and very severe RSV-LRTI) and total severe RSV-LRTI (severe LRTI and very severe LRTI) using the Cox proportional hazard model.
Adjusted HRs (aHRs) were calculated by adjusting for age, sex and place of living (municipality) and other factors with P < 0.1 in the univariate analysis. Socioeconomic status was assessed using the Simple Poverty Scorecard ® Poverty-Assessment Tool Philippines. 32 Extreme poverty was defined as a score <30 points. The hazard ratio for exclusive breastfeeding was calculated separately for children <6 months. The frequencies of the clinical manifestations were compared between children with severe (severe RSV-LRTI and very severe RSV-LRTI) and non-severe (RSV-RTI and RSV-LRTI) cases, using logistic regression models. Adjusted odds ratios (aORs) were calculated by adjusting for age, sex and place of living. The cox proportional hazard regression models and the logistic regression models were performed using R version 3.4.4, 33 with the survival package. 34 Figure S1). Among total severe RSV-LRTI (n = 101), 26 had SpO 2 measured after starting oxygen treatment and at least 24% (24/101) had SpO 2 < 90%, but only three were hospitalised. The proportion of RSV-associated hospitalisation was 7.1% (28/395). Of these (n = 28), SpO 2 data were available for 12 children and 25% (3/12) of those had SpO 2 < 90%. There were 13 undefined RSV-RTI episodes in children aged 0-1 month, of which eight (61.5%) were hospitalised, all with chest indrawing, and three (23.1%) had SpO 2 < 93% (Supporting Information Table S3). None of the 408 RSV-RTI episodes was fatal.

| Comparison between classification systems of severity
Among the 395 cases of children aged 2 months or older, using the RSV-LRTI severity definition, the median age of children with RSV-RTI was inversely associated with severity (Table 1). However, 71% (51/72) of the severe RSV-LRTI and 51% (15/29) of the very severe RSV-LRTI episodes occurred in children aged one year or older. The number of total RSV-RTI in the municipality of Caibiran was twice that of Kawayan. In RSV-associated hospitalisation, 96.4% (27/28) were classified as severe or very severe RSV-LRTI. One hospitalised child had chest indrawing and inability to feed; however, this child did not have either fast breathing or SpO 2 < 95%; therefore, this case was not classified as RSV-LRTI.
Within these 395 episodes, the subgroup of RSV was identified in 376 episodes. The proportion of severe or very severe episodes was higher in the subgroup A compared with the subgroup B without statistical significance (43.8% vs 29.9%, P = 0.06). Other viruses were also detected in 11.4% (46/395) of the episodes. Rhinovirus was the most frequently co-detected virus (30/46) across all severity categories of RSV-RTI (Table 1).

| Age-specific IRs of RSV-LRTI
The overall IRs of total RSV-LRTI and total severe RSV-LRTI were 62. 1 Table S4). Incidence rates for children aged 2-23 months were 124.0 and 51.5 per 1000 child-years for total RSV-LRTI and total severe RSV-LRTI, respectively, and these IRs were significantly higher than those of children aged 24-59 months (P < 0.001). Children aged 3-5 months had the highest IRs for total RSV-LRTI (207.4 per 1000 childyears) and total severe RSV-LRTI (74.5 per 1000 child-years). The IRs of total RSV-LRTI and total severe RSV-LRTI declined in children aged 6-8 months compared with children aged 3-5 months (P = 0.001 and P = 0.125, respectively). Total RSV-LRTI increased again in children aged 12-14 months compared with children aged 9-11 months without statistical significance (P = 0.07). After 20 months old, IRs declined with age. The same trends were seen when IRs were calculated by RSV subgroup, municipality or epidemics (Supporting Information Figure S2). When we considered RSV-RTI with ≥60 breaths/min or SpO 2 < 95% as LRTI and RSV-RTI with chest indrawing or SpO 2 < 93% as severe LRTI for children aged 0-1 month, 8 out of 13 were severe RSV-LRTI and the overall incidence rate of total RSV-LRTI and that of total severe RSV-LRTI were 63.1 and 23.5 per 1000 child-years in children aged 0-59 months, respectively.

| Risk factor analysis for developing total RSV-LRTI and total severe RSV-LRTI
In the univariate analysis, younger age group, place of living and educational level of caregiver were significantly associated with developing both total RSV-LRTI and total severe RSV-LRTI (Table 2).

| D ISCUSS I ON
Using data from a prospective cohort study of rural communities in the Philippines, we examined the incidence rates and severity of RSV-LRTI in detail. We estimated the overall IR of total RSV-LRTI to We found that age-specific IRs were significantly higher in children younger than 2 years old than in the older age group, and the same pattern was observed for total severe LRTI. Among those, the IRs of total RSV-LRTI and total severe RSV-LRTI were especially high in children aged 3-5 months. Peak IRs found in previous similar studies were 3-5 months in Nigeria, 0-5 months in Kenya and 6-8 months in Indonesia. 15,18,20 In the Indonesian study, the IR was lower in children aged 3-5 months than in those aged 6-8 months, which may have been related to different factors such as the level of maternal antibodies 35,36 or breastfeeding practice, 37 since young infants were likely to develop RSV-associated bronchiolitis and F I G U R E 2 Age-specific incidence rates of RSV-LRTI and distributions of followed up child-years in children aged less than 5 y in Biliran cohort study from April 2014 to March 2016. LRTI, lower respiratory tract illness; RSV, respiratory syncytial virus. The proportion of severe episodes was calculated as proportion of the total severe RSV-LRTI episodes among total RSV-LRTI by every 1 y of age. Incidence rates were calculated for 2 mo of age and every 3 mo of ages from 3 to 59 mo of ages. Child-years were calculated for every 3 mo of ages. The childyears during RSV epidemic period is shown as dark grey in background, and light grey shows child-years out of RSV epidemic period. pneumonia. 38 Interestingly, we found that IRs dropped in children aged 6-8 months and 9-11 months, then increased again in children aged 12-14, 15-17 and 18-20 months (Figure 2), in a pattern similar to previous studies. 15,18,20 In hospital-based studies, a majority of RSV-associated severe infections are seen in infants aged <1 year, especially in young infants aged <6 months. 19,39 However, our study and other prospective cohort studies have identified similar bimodal peaks in IRs of total RSV-LRTI, that is, young infants aged <6 months and children aged 12-23 months. We suspected that the increased IRs in children aged 12-20 months might be due to the change in the criteria for fast breathing at 12 months old. However, the agespecific IRs of severe RSV-LRTI defined by only chest indrawing or SpO 2 < 93% showed similar patterns, suggesting a true increase in incidence in the second years of life. Age-specific IRs might also be affected by the timing of RSV epidemics since observed child-year for each age group did not cover whole epidemic periods. Therefore, we compared the proportion of epidemic period in total observed child-years for each age group. We found not much difference in the proportions of epidemic periods between age groups. The agespecific pattern of IRs for RSV-LRTI has important implications for future intervention strategies including vaccination. While current vaccination strategies focus primarily on protecting young infants, if the IRs of RSV-LRTI (especially the more severe cases) are also high in children 12-23 months, vaccination strategies may need to consider including a booster dose before 12 months.
In our study, 13 undefined RSV-RTI episodes were identified in children aged <2 months due to unavailability of severity definitions. 11,26 Eight were hospitalised, and all of these had chest indrawing and three of them had SpO 2 < 93%. Therefore, more than half of RSV-RTI in children aged <2 months were considered to be severe.
Incidence rates for those with chest drawing and SpO 2 < 93% were highest in this age group, indicating the significant impact of RSV-RTI in this age group. To reveal true burden and risk factors for this age group, further studies, particularly large-scale birth cohort studies, should be conducted.
Total RSV-LRTI consists of RSV-LRTI, severe RSV-LRTI and very severe RSV-LRTI. Total severe RSV-LRTI consists of severe RSV-LRTI and very severe RSV-LRTI. Caregiver is the person that takes care of the child, knows best about the child and makes decisions for the best interest of the child. Solid fuel contains wood, charcoal, paper, etc a Adjusted for age group, sex, place of living, gestational age, birthweight, educational level of caregiver and number of other children <6 y in household.
b Adjusted for age group, sex, place of living, gestational age, birth order, educational level of caregiver and father and number of other children <6 y in household.
c Hazard ratio for breastfeeding status was adjusted for age group, sex, place of living and educational level of caregiver and calculated using subset data for 2-6 y old.    and non-exclusive breastfeeding (OR: 2.2), identified in a metaanalysis, 42 were not significant risk factors in the present study.
Prematurity and low birthweight were significantly associated with acute lower respiratory infection in case-control studies conducted in the hospital settings with a large number of children with severe RSV-LRTI in high-income countries. 42 In our study, which mainly included cases in primary care, we did not reach the statistical significance probably due to the insufficient number of children with low birthweight, prematurity and severe RSV-LRTI.
In other low-and middle-income countries, only unpublished data of one case-control study and one cohort study for prematurity and one cohort study for low birthweight were conducted and none of these studies showed prematurity or low birthweight as a significant risk factor for severe RSV-LRTI. 42 Children whose gestational age or birthweight was not reported by their mothers (categorised as "unknown") were less likely to develop RSV-LRTI compared with the reference group. There may be a recall bias that mothers who could not recall were more likely to have had children with normal weight and gestational age.
We included a relatively large number of children in our cohort study, but the number was still too small to assess all possible risk factors. Most other studies of risk factors have been conducted as casecontrol studies, which are more suitable for evaluating risk factors that occur with low frequency, such as prematurity and low birthweight. 42 In another prospective cohort study in Bohol, the Philippines, being male and having siblings were identified as significant risk factors, 43 although there were some differences in the study designs. The Bohol study used RSV-associated hospitalisation as an outcome and had twice as many participants as our study. On the other hand, a birth cohort study in Kilifi, Kenya, with fewer participants than ours, 44 having 1-2 siblings less than 6 years old living in the same household, was a statistically significant risk factor for developing total RSV-LRTI. Thus, further studies are necessary to comprehensively evaluate various risk factors for developing severe RSV infections.
As a previous study showed, 45  Taken together, testing for hypoxaemia is crucial for evaluation of children with RSV-associated RTI to identify severe cases. Home care for children with LRI should be considered with SpO 2 measurement. However, setting a cut-off point for application of the oxygen therapy and providing the necessary equipment are important to support health workers' decision.
One limitation of our study was that we only included episodes observed in health facilities. Since health-seeking behaviour can vary among age groups, 49 we may have underestimated the IRs, particularly in older children who are less likely to visit health facilities than infants. Also, our study only included two sites from the same small island with two epidemics over two years, and thus, our results, including IRs, may not reflect the true burden. In addition, the study was not conducted as a birth cohort and most newborn infants were not recruited for the study immediately after birth. Therefore, observed child-years in this age group were particularly small in infants aged less than 3 months.