Estimates of influenza and respiratory syncytial virus incidences with fraction modeling approach in Baguio City, the Philippines, 2012‐2014

Background Estimation of the incidences of influenza and respiratory syncytial virus (RSV) infection is important for disease control. Previous estimate in the city showed a substantial burden of influenza in both outpatients and inpatients while it did not account for individuals who do not seek medical attention nor RSV. Patients/Methods A total of 17 674 influenza‐like illness (ILI) and 13 242 severe acute respiratory illness (SARI) cases were recruited, and samples were collected from 6267 and 2962 of ILI and SARI cases, respectively. RT‐PCR assays were performed to detect influenza and RSV in the samples. A health‐seeking behavior survey was conducted from February 2014 to April 2014 to estimate the fraction of infected individuals who did not seek medical attention between rainy and dry season. Results Average influenza and RSV incidence rates in outpatients were 1.6 and 1.4 per 1000 individuals, respectively, and the highest incidence rate for both viruses was found in the of 6‐23 month age group. Average influenza and RSV hospitalization incidence rates were 1.7 and 1.9 per 1000 individuals, respectively. Further, we estimated that the incidence rates of influenza and RSV in individuals who did not seek medical attention were threefold and 1.6‐fold those in the medically attended population. Conclusions Respiratory syncytial virus and influenza pose a substantial disease burden, particularly in hospitalized cases. The implementation of either a community‐based approach or an enhanced surveillance system in combination with a community survey will allow a better understanding of the disease burdens of RSV and influenza in the Philippines.

of this work, there remains an important gap in estimating influenza incidence in the Philippines, largely owing to the presence of an influenza-infected population that does not seek medical attention during their illness.
Respiratory syncytial virus (RSV) causes acute respiratory illness, particularly in young children. 4 Estimates of RSV incidence are highly variable across study populations, 5 and the incidence of this disease is sparsely reported, particularly in developing countries in tropical regions. 6,7 Currently some vaccines for RSV are in the trial phase, and RSV incidence rate is one of the primary endpoints being used to evaluate the effectiveness of these vaccines. 5 Additionally, estimates of the RSV incidence rate provide useful information for prioritizing the implementation of public health services within a country.
This study aimed to estimate the incidence rates of influenza and RSV in Baguio City as well as to compare estimates of non-medically attended ILI episodes with health-seeking behavior data.

| Study sites
Baguio City is the regional center of the Cordillera Administrative Region, which is located in the northern part of Luzon Island, and has a population of 315 000. In the city, there are 16 public health centers (PHCs) providing a variety of primary medical services including immunization and morbidity consultations to inhabitants in the catchment areas. In addition to the PHCs, five hospitals and private clinics are major facilities to provide medical services.

| ILI and SARI surveillance
The details of ILI and SARI surveillance in the Philippines have been previously documented. 3 Briefly, an ILI case is defined as a patient who developed a sudden onset of a fever (>38°C) with a cough or sore throat. A SARI case is defined as a patient who developed shortness of breath or difficulty in breathing that required hospital admission. Children who were diagnosed with cases of pneumonia or severe pneumonia according to the integrated management of childhood illness guidelines 8 were also enrolled in our study. ILI and SARI surveillance are operated in all PHCs and hospitals in the city.
A structured questionnaire was used, and naso-or oropharyngeal swabs were collected by field nurses on each surveillance day. The surveillance day was one of the two morbidity consultation days in PHCs. Field nurses visited to hospital once a week to collect specimens from SARI cases if the interval days from onset did not excess to 5 days. The weekly numbers of ILI and SARI cases who visited on days other than these surveillance days were then counted. All collected specimens were tested for influenza A and its subtypes (hereafter, A(H1)pdm09 or A(H3)) by performing RT-PCR assays 9 or tested for influenza B and RSV by performing conventional RT-PCR assays. 10 The study period was between 1 January 2012 and 31 December 2014.

| Health-seeking behavior survey
Each PHC has a catchment area which consists of several barangays (local administrative division) for their health services. All PHCs were categorized into three groups according to ILI consultation activity level (low, moderate, or high) and into two groups by their shortest distance from city hall (suburb or city center). The consultation activity level was defined by the average number of annual ILI consultations reported in 2012 and 2013. All PHCs were allocated in the above 2×3 matrix (2-3 PHCs in each cell), and a PHC was each randomly selected from the cell (totally 6 PHCs). Two or three barangays were then randomly selected for each PHC covering area. The total sample size was calculated as 2800 based on the assumption that 50% of the population 11 would have sought medical consultation with a precision of 5%. A design effect of two was taken into account owing to the clustering of cases by barangay and household.
The sample design in the survey was a self-weighting. The number of study population was proportionally allocated by population size in each barangay and also by the age structure of population in the city. The survey was conducted between February 2014 and April 2014. All interviews were conducted with a person of targeted age group by trained interviewers, and a field coordinator supervised their activities. A standardized questionnaire was used to obtain respondents' demographic and socioeconomic information, household structure, and an attitude as well as practice to seek medical consultation after developing ILI-and SARI-like symptoms (Supplement information). Symptoms in case definitions of both ILI and SARI were explained during the interview. We asked all participants about experiences of both ILI and SARI episodes 2 months prior to the survey.
We then estimated frequencies of different health-seeking behavior in ILI and SARI episodes. The survey was conducted during dry season in the city.

| Statistical analysis
The virus-positivity rate was calculated as the proportion of the number of positive samples to the total number of tested samples. The population was projected based on the 2010 census data. We estimated the influenza incidence rate by each age group, as expressed per 1000 individuals, by dividing the number of patients with ILI who presented at the health facilities and whose samples tested positive for influenza by the census population. Then, we multiplied the result with the inverse of the proportion of the number of samples obtained to the total ILI count. A similar procedure was used for estimating RSV-and influenza-associated SARI incidence rates. We estimated the overall influenza incidence rate based on cases of A(H1)pdm09, A(H3), and influenza B. A 95% confidence interval (CI) was estimated using simple exact binomial CIs. A Bayesian regression model was constructed to estimate the number of ILI cases who sought care at health facilities other than PHCs from the number of cases who visited PHCs: where Y i,j,k and X i,j,k are the number of ILI cases during the rainy season and dry seasons, respectively, at age group i, category j, and distance k. C i,j and S i,j are the rate of PHC consultation activity and the rate of intent to seek consultations during the rainy season (June to December), respectively. The fraction of non-medically attended to medically attended cases was then applied to obtain the total number of ILI consultations in each PHC area. All analyses were performed in R version 3.1.0, 12 and rstan version 2.8.0 was used to construct the model. 13

| Ethical issues
Written informed consent was obtained prior to participation, primar-

| Virus-positive samples among ILI and SARI cases
In total, 17 674 ILI and 13 242 SARI cases were recruited throughout the study period. The number of both ILI and SARI cases was not different between rainy and dry seasons (P=.99 respectively). Of those, samples were collected from 6267 ILI cases (35.5%) and 2962 SARI cases (22.4%). When comparing the ages of the ILI and SARI patients, we found that children accounted for the highest number of cases with ILI, while the highest number of SARI cases was in adults over 50 years of age ( Table 1). Fifteen of the SARI cases were fatal; however, except for one case each with influenza B and RSV, who both returned home against medical advice, all of these patients tested negative for the targeted viruses. For the ILI cases, influenza-positive samples were detected during 84.6% of the study weeks, while RSVpositive samples were only detected during 53.9% of the study weeks ( Figure 1).
The virus with the highest number of positives varied across the study years (Table 2) respectively. The highest numbers of A(H1)pdm09 and A(H3) cases were both found among the age group of 2-4 years. Of the ILI and SARI cases whose samples were collected, 2.4% and 9.4%, respectively, had received an influenza vaccination within a year prior to their illness. Within these subsets, five A(H3) and 11 influenza B cases were detected among the ILI cases, while 11 A(H3), three A(H1)pdm09, and 20 influenza B cases were detected among the SARI cases. A(H3)-and influenza B-positive cases in the SARI group were found most commonly in the <2 year age group.
We estimated the incidence rates of influenza and RSV among the ILI cases by study year and age group (Table 3). Overall, the average influenza and RSV incidence rates were 1.6 and 1.4 per 1000 individuals, respectively, and the highest incidence rate for both viruses was found in the of 6-23 month age group. Interestingly, the ratio of incidence rates in the of 2-4 year age group to that in the 6-23 month age group was significantly lower for RSV than that for influenza (0.38 and 0.62, respectively). For both RSV and influenza, children aged <6 months generally had a lower incidence rate than either of those two groups.
This trend is more distinct for the RSV incidence rate, while the influenza incidence rates in these three groups were more varied among study years and virus types.
For SARI cases, the average influenza and RSV hospitalization incidence rates were 1.7 and 1.9 per 1000 individuals, respectively.
Notably, although the overall hospitalization impacts of influenza and RSV were similar to one another, the age-stratified incidences were remarkably different (

| Health-seeking behavior survey results
In total, 2655 respondents were enrolled in the health-seeking behavior survey. The median age of the respondents was 10 years old (range: 1 month-100 years of age), and the male-to-female ratio was

| Adjusted incidence rates for influenza and RSV
We adjusted our estimates of the influenza and RSV incidence rates based on the size of the residences of patients who did not visit PHCs, PHC consultation activity, and differences in health-seeking behaviors during the rainy and dry seasons ( Table 6). These adjusted influenza and RSV incidence rates were estimated as 4.8 and 2.3 per 1000 individuals, respectively. Notably, after these adjustments, the RSV incidence rate was remarkably high in children <6 months of age. The trends in the adjusted incidence rates of both influenza and RSV by age group are similar to the unadjusted incidence rates, which did not account for the group of infected individuals who did not seek medical attention; however, the adjusted incidence rates in adults >50 years were higher than the unadjusted incidence rates.

| DISCUSSION
One global study estimated that there is a substantial influenza burden in children aged <5 years and in school-aged children. 14  Respiratory syncytial virus is a major etiological agent for both ILI and SARI cases among young children. 17,18 Although the overall RSV incidence rates of outpatients and inpatients in this study were comparable to those of influenza, we found that there was a higher incidence rate in young children for RSV than for influenza, which is in agreement with reports from previous studies. 7,19 Because the Philippines has a population structure with a high percentage of children, the burden of RSV is relatively high. There were very few fatalities among the study participants whose samples were determined to be positive for influenza or RSV. This is likely due in part to the early treatment that the study participants received, as three-quarters of the subjects sought medical attention within 2 days. The small proportion of elderly individuals >65 years (3.5% of the total population) in our study, who typically contribute heavily to both influenza and RSV mortality, [20][21][22] may have also contributed to the low number of fatalities in our study.
We estimated the incidence rates of influenza and RSV as threefold and 1.6-fold, respectively, of the rates in the medically attended population. This is largely reflective of the size of the adolescent and adult population who did not seek medical care according to the health-seeking behavior survey. These estimates are slightly lower than the rates estimated by a similar study conducted in Kenya. 7 A surveillance system that targets the medically attended population can provide a solid range of incidence estimates for both individuals who receive medical attention and those who do not; however, the nature of such systems inherently includes the potential for underestimating overall incidence. Influenza episodes that do not receive medical attention lead to fewer direct economic losses than medically attended cases, but they impact indirect costs, such as productivity costs. 23 Therefore, it is important to explore disease burden estimates in terms of both direct and indirect costs. Compared with the difference in the incidence rate between medically and non-medically attended groups for influenza, less of a difference in incidence rates between these groups was observed for RSV in our study and the  ILI, influenza-like illness; SARI, severe acute respiratory illness; A(H1)pdm09, influenza A(H1N1)pdm09; A(H3), influenza A(H3N2); RSV, respiratory syncytial virus.
Kenyan study. The relatively young population that comprises the peak age group for RSV infection seeks medical care more frequently than the rest of the population.
We also found a difference in health-seeking behavior between the rainy and dry seasons. Health-seeking behavior has been widely used to calculate the function of the proportional range between  medically attended and non-medically attended groups. 7,24,25 The results of a single cross-sectional survey may better reflect recent episodes in respondents. That is, a question structure that considers the effect of a seasonal pattern may grasp behavioral attitudes more precisely than one that does not consider this factor.
There are some limitations in our study. First, owing to limited resources, we could not obtain samples from the ILI who visited facilities other than sampling days and from SARI cases whose onset was ≥5 days before sampling days, and this may have potentially lead to underestimates of the influenza and RSV incidence rates in both the medically attended and the non-medically attended populations.
Second, although we used the results of our health-seeking behavior survey, which accounted for seasons as well as the levels of PHC consultation activity, to calculate a fraction of non-medically attended individuals, these data were produced from a single survey. Although there was no significance difference in terms of the number of ILI and SARI cases between rainy and dry season, influenza A(H3) was predominantly circulated and both influenza B and RSV were less found.
A variety of circulating viruses may affect the fraction of medically attended to non-medically attended cases. We calculated the fraction of season by multiplying a ratio of proportions of medically attended between rainy and dry seasons to frequencies of seeking facilities in last 2 months prior to the interview conducted in dry season because we intended to minimize the effect of recall bias. A subsequent health-seeking behavior survey in rainy season could estimate above fraction more precisely. Despite these limitations, this study provides incidence rates of both influenza and RSV among inpatients and outpatients that take into account the non-medically attended population in the Philippines.
Respiratory syncytial virus and influenza pose a substantial disease burden in the city. To minimize this impact, particularly that of hospitalization, control measures, such as influenza vaccinations, 26 should be considered. 27 Additionally, the current facility-based surveillance systems may underestimate the influenza incidence in  infected individuals who do not seek medical attention. The implementation of either a community-based approach or an enhanced surveillance system in combination with a community survey will allow a better understanding of the disease burdens of RSV and influenza in the Philippines.