Estimating burden of influenza‐associated influenza‐like illness and severe acute respiratory infection at public healthcare facilities in Romania during the 2011/12‐2015/16 influenza seasons

Background Influenza is responsible for substantial morbidity and mortality, but there is limited information on reliable disease burden estimates, especially from middle‐income countries in the WHO European Region. Objectives To estimate the incidence of medically attended influenza‐associated influenza‐like illness (ILI) and hospitalizations due to severe acute respiratory infection (SARI) presenting to public healthcare facilities in Romania. Patients/Methods Sentinel influenza surveillance data for ILI and SARI from 2011/12‐2015/16, including virological data, were used to estimate influenza‐associated ILI and SARI incidence/100 000 and their 95% confidence intervals (95% CI). Results The overall annual incidence of ILI and influenza‐associated ILI per 100 000 persons in Romania varied between 68 (95% CI: 61‐76) and 318 (95% CI: 298‐338) and between 23 (95% CI: 19‐29) and 189 (95% CI: 149‐240), respectively. The highest ILI and influenza incidence was among children aged 0‐4 years. We estimated that SARI incidence per 100 000 persons was 6 (95% CI: 5‐7) to 9 (95% CI: 8‐10), of which 2 (95% CI: 1‐2) to 3 (95% CI: 2‐4) were due to influenza. Up to 0.3% of the Romanian population were annually reported with ILI, and 0.01% was hospitalized with SARI, of which as much as one‐third could be explained by influenza. Conclusions This evaluation was the first study estimating influenza burden in Romania. We found that during each influenza season, a substantial number of persons in Romania suffer from influenza‐related ILI or are hospitalized due to influenza‐associated SARI.


| INTRODUCTION
Influenza is a respiratory disease responsible for substantial increase in morbidity, mortality and costs during seasonal epidemics and pandemics. [1][2][3][4][5][6] The World Health Organization (WHO) estimates that annually seasonal influenza epidemics result in an estimated 3-5 million cases of severe disease and 250 000-500 00 deaths. 7 However, the burden of seasonal influenza can vary between influenza seasons and countries. Therefore, reliable disease burden estimates are needed, especially from low-and middle-income countries to provide a better understanding of the impact of influenza. Knowledge about influenza burden is crucial to further develop and strengthen influenza prevention strategies, as well as to assess utility. Thus, national estimates would enable governments, non-governmental agencies, and others to make informed evidence-based decisions when allocating limited resources dedicated to public health and planning intervention strategies to limit the spread of influenza.
While some estimates of the national burden of influenza are available for countries of the WHO European Region, the majority of primary study estimates come from Western Europe and no estimates are available for over half of the member states. 8,9 Except for a study on the economic influenza burden among the elderly in 6 middle-income countries of Central and Eastern Europe in 2010/11, 9 influenza burden in the Romanian population has never been estimated.
To gain insight into medically attended influenza burden on a national level, we performed an assessment of influenza disease burden due to influenza-like illness (ILI) attendances and severe acute respiratory infection (SARI) hospitalizations associated with influenza in Romania at public facilities for the 2011/12-2015/16 influenza seasons.

| METHODS
This study followed the methods outlined in the WHO Manual, A

Manual for Estimating Disease Burden Associated with Seasonal
Influenza in a Population, using 5 years of Romania national surveillance data. 10  These GP practices were selected by the counties' public health authorities and agreed to participate. Since 2014, sentinel ILI surveillance is performed in 15 counties and Bucharest and included at least 192 GPs.
Cases were identified from patients who sought care at the clinic and those receiving home visits. In line with the WHO definition, ILI was defined as an acute respiratory infection with measured fever of ≥38°C and cough, with onset within the last 10 days. 12 Samples for virological testing were collected 1 day per week (first two cases at each participating GP practice every Tuesday). Samples were only collected from ILI cases belonging to groups targeted for vaccination, including pregnant women, persons with underlying chronic conditions, aged >=65 years, HCWs, residents of long-term care facilities. In Romania, the majority of the population is covered by a universal health insurance and registered with a GP practice. As these patients must seek care from the GPs with which they are registered, the catchment population was estimated by summing the patients registered at each participating GP practice.

| Severe acute respiratory infection sentinel surveillance and catchment population
Since 2009, hospital-based sentinel surveillance for SARI in Romania has been conducted in 20 university hospitals in 7 counties from all 4 geographical regions ( Figure 1).
A SARI case was defined as an acute respiratory infection with an onset in the previous 7 days requiring overnight hospitalization, history of fever or measured fever ≥38°C, cough and shortness of breath or difficulty breathing. 11,12 Starting with the 2013/14 season, the case definition was expanded to include an onset within the previous 10 days. Information on pregnancy status, underlying medical conditions and the outcome of the illness was also recorded for each SARI case. A subset of SARI cases was tested for influenza virus using RT-PCR. The sampling algorithm for SARI cases had a two-step approach with greater sampling early in the influenza season and decreased once influenza circulation had been identified. Starting in week 46, the first three identified SARI cases in each sentinel hospital were sampled per week. This sampling strategy continued until ≥10% of all the sentinel surveillance system samples (ILI and SARI) tested positive for the same influenza subtype/variant. Once the positivity rate by subtype/variant reached 10%, only the first identified SARI case in each sentinel hospital each week was sampled for the remainder of the influenza season.
Because the catchment population for each hospital was unknown, an estimate was made by dividing the total number of all-cause hospitalizations for each sentinel hospital by the total discharges (including deaths) from all the public hospitals in Romania to determine the proportion of all-cause discharges for each sentinel hospital. This proportion was used as a proxy for the proportion of the population that sought care at each hospital and it was then multiplied by the Romanian population for each year to determine the catchment population for each sentinel site. The catchment population for all sites was summed to produce the estimate of the catchment population of all sentinel hospitals.

| Statistical analysis
To estimate the annual incidence of ILI and SARI by type and subtype, we summed the weekly incidence in the sentinel sites of each influenza season and calculated their 95% confidence intervals by 1) calculating the error factor (EF), which was the exponentiation of 1.96 divided by the square root of number of (influenza-associated) ILI or SARI cases, and 2) calculating the range of the 95% CI by dividing (higher CI) or multiplying (lower CI) the incidence by the EF, as described in WHO Manual. 10 We stratified the ILI and influenza-associated ILI incidence by age into 0-4 years, 5-14 years, 15-64 years and 65 years and older groups. Due to limited availability of the hospital discharge data by different age groups, we were only able to stratify SARI and laboratory-confirmed influenza-associated SARI cases into paediatric (<16 years) and adult (≥16 years) age groups. All incidences were extrapolated to the Romanian population.
We calculated case fatality proportion (CFP, %) among laboratoryconfirmed influenza-associated SARI cases by dividing the number of laboratory-confirmed influenza-associated deaths among the SARI patients by the total laboratory-confirmed influenza-associated SARI cases and multiplying it by 100%. We estimated its 95% confidence interval by the Clopper-Pearson method. 13 The Romanian population size by age groups was extracted from the National Institute of Statistics in Romania as of January 1 for the years 2011-2016 (http://www.insse.ro/cms/en). Approximately, 5% of the Romanian population was 0-4 years old, 11% was 5-14 years old, 68% was 15-64 years old, and 16% was 65 years and older.

| Ethical issues
The project was reviewed by the Ministry of Health of Romania and judged to be a public health evaluation. Only anonymized records were used for the analysis.

| RESULTS
During the study period from the 2011/2012-2015/2016, Romania experienced three seasonal influenza epidemics predominately due to influenza A virus and two seasons due to both influenza A and B (Table 1). One season was dominated by influenza A(H3) and two seasons by influenza A(H1N1)pdm09.
The incidence of both ILI and influenza-associated ILI was highest in children, especially children under 5 years of age (Table 2), where approximately 20% of ILI cases occurred ( Table 1) (Table 2).
On average, approximately 27%-56% of ILI cases in Romania could be explained by influenza (Table 1), and positivity among the SARI cases also varied over influenza seasons 24%-43% (Table 3). Influenzaassociated SARI incidence was higher among the <16-year-old cases, It was not possible to calculate the incidence for these groups as the catchment population for these target groups was not known.

| DISCUSSION
Our results suggest that influenza is associated with substantial morbidity in terms of primary care attendances due to ILI and hospitalizations T A B L E 1 Influenza by (sub)type, ILI and catchment population of the    Using the Romanian influenza sentinel system, we were able to estimate clinical burden of ILI and SARI, including the proportion attributable to influenza. In order to compare influenza disease burden to other countries in the region, we therefore used methodology proposed by the WHO 10 to estimate influenza burden in Romania.

| Limitations
This study sought to estimate the burden of influenza at public healthcare facilities in Romania. Persons who went to private clinics T A B L E 4 Incidence of SARI and influenza-associated SARI by virus (sub)type and paediatric and adult age group per 100 000 population in Romania in 2011/12-2015/16 influenza seasons This study relied on data from the national influenza sentinel surveillance system, which has been collecting ILI and SARI data in Romania since 2009. These data were collected for public health purposes relying on data voluntarily collected from clinicians to detect and report cases in addition to their normal duties, and may be limited in comparison with data collected in targeted research programmes. Although steps were taken to minimize potential bias, the surveillance system consisted of many doctors from numerous sites.
Differences in adherence to protocols, hospitalization practices and patient healthcare-seeking behaviour could not be accounted for in the analysis.
Samples for influenza testing for ILI cases were only collected from patients who belonged to the risk groups targeted for vaccination in Romania. In this analysis, we used the per cent of the influenzapositive ILI cases from these vaccine target groups to estimate the influenza positivity for all ILI cases. If these cases from the vaccine target groups were more likely to be influenza positive on testing, then we may have overestimated the ILI rates. Unfortunately, it was not possible to determine the incidence in the risk groups as the Romanian population for each risk group was not known. Also, the number of pregnant women in the surveillance sites was very low, so we were not able to estimate influenza burden in this specific group. Although influenza has been shown to cause severe disease at higher incidence in the very young and the elderly, 7 we were not able to examine the incidence in these age groups. We were only able to calculate the SARI incidence for children less than 16 years of age and all adults (≥16 years old), because hospital discharge data were only available for these groups. In addition, very low vaccination uptake rates among the cases registered by the surveillance system did not allow us to estimate influenza burden between the vaccinated and the unvaccinated individuals.
During the study period, two changes were made to the sen- sampled. This may not have been a sufficient number to adequately represent the relative intensity of influenza activity and to refine per cent positive influenza cases needed for estimation of the influenzaassociated ILI and SARI in Romania, especially during the height of the influenza season when the relative sampling fraction is lowest.
A systematic evaluation of the national sentinel surveillance system would provide information on adherence to case definitions and protocol, optimization of sampling strategy to meet system objectives and system quality. This enhanced understanding of the sentinel surveillance system data would assist when future studies are performed to assess influenza burden in the groups targeted for influenza vaccination in Romania and the cost-effectiveness of influenza vaccination.

| CONCLUSION
This evaluation using the recently proposed methodology by WHO 10 was the first study estimating influenza burden in Romania. We found that during each influenza season, a substantial number of persons in Romania suffer from influenza-related ILI or are hospitalized due to influenza-associated SARI.
This information will be used to support influenza prevention campaigns during the influenza season in Romania to improve vaccination uptake and increase awareness of influenza prevention and treatment.

ACKNOWLEDGEMENT
We would like to thank the staff participating in influenza surveillance system in Romania.