First-year results of the Global Influenza Hospital Surveillance Network: 2012–2013 Northern hemisphere influenza season

Background The Global Influenza Hospital Surveillance Network (GIHSN) was developed to improve understanding of severe influenza infection, as represented by hospitalized cases. The GIHSN is composed of coordinating sites, mainly affiliated with health authorities, each of which supervises and compiles data from one to seven hospitals. This report describes the distribution of influenza viruses A(H1N1), A(H3N2), B/Victoria, and B/Yamagata resulting in hospitalization during 2012–2013, the network’s first year. Methods In 2012–2013, the GIHSN included 21 hospitals (five in Spain, five in France, four in the Russian Federation, and seven in Turkey). All hospitals used a reference protocol and core questionnaire to collect data, and data were consolidated at five coordinating sites. Influenza infection was confirmed by reverse-transcription polymerase chain reaction. Hospitalized patients admitted within 7 days of onset of influenza-like illness were included in the analysis. Results Of 5034 patients included with polymerase chain reaction results, 1545 (30.7%) were positive for influenza. Influenza A(H1N1), A(H3N2), and both B lineages co-circulated, although distributions varied greatly between coordinating sites and over time. All age groups were affected. A(H1N1) was the most common influenza strain isolated among hospitalized adults 18–64 years of age at four of five coordinating sites, whereas A(H3N2) and B viruses were isolated more often than A(H1N1) in adults ≥65 years of age at all five coordinating sites. A total of 16 deaths and 20 intensive care unit admissions were recorded among patients with influenza. Conclusions Influenza strains resulting in hospitalization varied greatly between coordinating sites and over time. These first-year results of the GIHSN are relevant, useful, and timely. Due to its broad regional representativeness and sustainable framework, this growing network should contribute substantially to understanding the epidemiology of influenza, particularly for more severe disease.


Background
According to the World Health Organization (WHO), seasonal influenza epidemics affect an estimated 5-15% of the total population worldwide, with 3-5 million cases of severe illness, resulting in 250,000-500,000 deaths [1]. However, few data are available for many parts of the world where active surveillance is lacking. In addition, the viruses and the severity of influenza epidemics vary greatly between years and geographical areas [2][3][4]. To address the rapidly evolving antigenicity of circulating influenza viruses, twice annually, the WHO re-evaluates the viruses that should be included in the seasonal influenza vaccines. To inform policy decisions, national health authorities need to understand the burden of influenza disease and the impact of current vaccination programs in their countries.
High-quality, active surveillance networks are needed to better understand influenza epidemiology and therefore better control influenza epidemics [5][6][7]. Data from existing sentinel physician networks are used in several countries to conduct annual studies on the effectiveness of vaccines in preventing medically attended influenzalike illness (ILI) [8][9][10][11][12]. These networks, however, do not collect data on the impact of influenza infection on hospitalization or on the impact of influenza vaccines on influenza-related hospitalization, which substantially influence evaluation of the benefits and cost-effectiveness of influenza vaccines [13].
Active surveillance networks are also powerful advocacy instruments for highlighting the often-underestimated impact of influenza [5]. While hospital surveillance systems already exist for detecting outbreaks of respiratory viruses [14][15][16], few focus on the actual burden of serious influenza cases using the specific outcome of laboratoryconfirmed influenza; instead, the burden is most often estimated from hospital databases using criteria prone to various biases [13,17].
The Global Influenza Hospital Surveillance Network (GIHSN) was initiated in 2011 to fill this gap in epidemiology and public health knowledge. The GIHSN is a public-private partnership between Sanofi Pasteur, FISABIO-Salud Pública, and several coordinating sites affiliated with national health authorities. In accordance with WHO recommendations [7], coordinating sites are selected based on their motivation, geographic representativeness, ability to conduct epidemiological studies, availability of laboratory facilities, and experience in influenza surveillance. Each coordinating site supervises a group of one to seven hospitals in its country or geographical region and follows a core reference protocol. The GIHSN has three main objectives: (i) evaluate the burden of severe influenza disease, defined as hospitalization related to community-acquired influenza or complications following an influenza infection; (ii) quantify the distribution of the different influenza viruses (A(H1N1), A (H3N2), B/Yamagata, and B/Victoria) among these severe cases; and (iii) measure the effectiveness of influenza seasonal vaccines to prevent these hospitalizations using a test-negative design.
In this report, we evaluated the characteristics of hospitalizations related to influenza and the temporal and geographic distribution of the different influenza viruses in these cases during the 2012-2013 Northern hemisphere influenza season, the program's first year.

Study design
This was a multi-centre, prospective, active surveillance, hospital-based epidemiological study during the 2012-2013 influenza season in 21 hospitals in Spain, France, Turkey, and the Russian Federation (Table 1). Data on hospitalized patients with a diagnosis possibly associated with influenza were collected by an active surveillance system composed of healthcare professionals trained to follow a common reference protocol. Data were consolidated at five coordinating sites, including one in Spain (Valencia), one in France, one in Turkey, and two in the Russian Federation (Moscow and St. Petersburg).
The reference protocol was adapted by the coordinating sites according to their local conditions. In particular, case identification was adapted to the specific local setting because of the differences in health care delivery systems between the different countries and differences in the types of hospitals included, although admission diagnosis codes for inclusion were the same for all sites. At each site, except Turkey, the study start was predefined according to the site experience of the influenza epidemic wave [8]. No study period was defined for Turkey because it was included as a pilot study. Specificities between coordinating sites in the application of the reference protocol are summarized in Table 2

Study population
Non-institutionalized patients hospitalized for at least 24 h with a diagnosis possibly associated with influenza were considered eligible to be included in the study (see Additional file 1: Table S1 for admission diagnosis codes). Patients in the Russian Federation, Spain, and Turkey could be of any age, whereas in France, only patients ≥18 years of age were screened.
Subjects ≥5 years of age had to meet the European Centre for Disease Prevention and Control's clinical case definition of influenza-like-illness (ILI) [18], which included at least one of four systemic symptoms (fever or feverishness, headache, myalgia, or malaise) and at least one of three respiratory symptoms (cough, sore throat or shortness of breath), although we did not include sudden onset of symptoms as a criterion. Subjects ≥5 years of age also had to have been hospitalized within 7 days of the onset of ILI. Children <5 years of age had to have been admitted to the hospital within 7 days of the appearance of symptoms potentially associated with influenza. Patients were excluded if they had been discharged from a hospital within 30 days of the current admission.

Data and sample collection
A nasopharyngeal swab was obtained from all patients. An additional pharyngeal swab was collected for patients ≥14 years of age and a nasal sample for children <14 years old. Samples were collected within 48 h of hospital admission and were stored at −20°C at the study site or were sent directly to the coordinating site's laboratory for testing. Influenza infection status, patient demographics, and influenza vaccination status were recorded with a core questionnaire via a combination of face-to-face interview of patients or legal representatives, interviews of attending physicians, and a review of clinical records.

Data management and statistical analysis
Coordinating sites collected anonymized data from the core questionnaires and checked for missing, inconsistent, or incorrect data. Whenever possible, queries of any inconsistencies or missing data were resolved by the investigators at each of the coordinating sites. Missing data were not replaced for the statistical analyses. Data from each coordinating site were shared with the network coordinating centre (FISABIO-Salud Pública, Valencia, Spain) via a secured, internet-based system.
Only samples taken within 7 days of symptom onset were included in the analysis. Descriptive statistics were calculated for the characteristics of influenza-associated hospitalization by age group, influenza virus and sociodemographic characteristics.
Statistical analysis and data management was performed using STATA version 12 (StataCorp, College Station, TX).

Results
A total of 8795 patients hospitalized for diagnoses possibly related to influenza were considered eligible for this study (

Epidemiology of hospitalization with laboratoryconfirmed influenza
Cases hospitalized with confirmed influenza peaked between the fourth and eighth epidemiological week of 2013 ( Figure 1). The timing of suspected influenza cases All four influenza viruses examined in this study (A (H1N1), A(H3N2), B/Yamagata, and B/Victoria) contributed to hospitalization with influenza at all five coordinating sites. The timing and distribution of viruses varied considerably between the coordinating sites ( Figure 2). In Moscow, A(H1N1) was the most frequent cause of confirmed influenza (53.7% of cases), peaking at epidemiological week 5 of 2013, followed by an increase in B and A(H3N2) towards the end of the season. A(H1N1) was the most frequent cause of confirmed influenza in Turkey, although results for the full season were not available. In St. Petersburg, A(H1N1) was a frequent cause of confirmed influenza (33.4%), A(H3N2) (20.2%) was also common and B lineages (37.9% overall) appeared later. In Spain, B/Yamagata was the most frequent (60.6% overall) and peaked around epidemiological week 6 of 2013, A(H1N1) increased later in the season. In France, the frequency of B virus (44.0% overall) was slightly higher than of the other influenza viruses, although it cocirculated with the influenza A viruses. Both B virus lineages circulated in all countries. The lineage included in the vaccine (B/Yamagata) was most common, representing 89.2% (436/489) of all B viruses cases in which the lineage was identified.
At all coordinating sites except for France, A(H1N1) was the most common influenza strain isolated from hospitalized adults 18-64 years of age. In contrast, A (H3N2) and B viruses were more common than A (H1N1) in elderly adults (≥65 years of age) at all coordinating sites (Table 4).

Demographics and clinical features of hospitalized laboratory-confirmed influenza cases
In Spain, France, and Turkey, more than half of influenza cases had one or more comorbidities, whereas at both coordinating sites in the Russian Federation, more than 80% were reported to have no comorbidities (Table 5). Cardiovascular disease, respiratory disease, and diabetes were the most frequent comorbidities at all coordinating sites among positives for influenza.   Values are n (%). Patients may have had more than one influenza virus due to a mixed infection.
In Spain and France, more than half of the subjects with influenza were ≥65 years of age, whereas <5% in St. Petersburg and Moscow were in this age group. Most of the subjects with influenza in Moscow were 18-64 years of age, while in St. Petersburg, the only site including dedicated paediatric hospitals, most were <18 years of age. The male-tofemale ratio for confirmed influenza cases was close to 1 in Spain, St. Petersburg, and France, whereas in Moscow, more women than men were included. Indeed, a maternity ward was included in this site, where 242 pregnant women had influenza. Amongst women 15 to 45 years of age hospitalized for diagnoses possibly related to influenza, pregnancy was significantly associated with a risk of confirmed influenza (adjusted OR, 1.36 [95% CI, 1.03 -1.80]).
A total of 16 deaths and 20 intensive care unit admissions were recorded. Outcomes were similar at the different coordinating sites, except for Turkey, which had a disproportionately high number of serious outcomes (11 intensive care unit admissions and 7 deaths). For the 16 deaths, the mean age was 60 years (range, 1 -93 years). Nine of these patients were >65 years of age, two had been vaccinated against seasonal influenza during the current season, and seven had more than one comorbidity.
At all five coordinating sites, the length of hospital stay was similar for all influenza circulating viruses (Additional file 3: Figure S1).

Influenza vaccine uptake
Influenza vaccination coverage among all hospitalized patients included in the study, whether positive or negative for influenza, was low in St. Petersburg (1.4%), Moscow (1.7%), and Turkey (14.5%) and moderate in Spain (38.9%) and France (47.0%). Influenza vaccination among all patients was highest in elderly adults in Spain (60.7%), Turkey (23.1%), and France (63.8%). In hospitalized children, regardless RT-PCR results, uptake rates at all coordinating sites were very low (≤5%).

Discussion
These results represent the first year's experience from the GIHSN, a prospective hospital-based influenza surveillance network. The data presented are from the 2012-2013 Northern hemisphere influenza season and were collected from 21 hospitals via five coordinating sites in four countries. Of a total of 6033 eligible patients, valid samples were obtained from 5034, of whom 1545 (30.7%) were positive for influenza. The study confirmed that influenza can result in serious outcomes not only in elderly people and those with comorbidities but also in the wider population, irrespective of age or sex. It also confirmed that pregnant women are at significant risk for hospitalization with influenza [19,20], which supports global recommendations for prioritizing influenza vaccination for pregnant women [21,22].
The study allowed us to examine the influenza viruses resulting in hospitalization. We found substantial heterogeneity in virus circulation between coordinating sites, even those in neighbouring geographical areas and even within the same site over time. Unexpectedly, influenza B, mostly the Yamagata lineage, was the most common virus identified in elderly adults, where it accounted for more than half of the hospitalized confirmed influenza cases. Thus, as found in a recent structured review [23], influenza B poses a significant health burden. Indeed, influenza B is now receiving increased attention, and quadrivalent influenza vaccines containing both B lineages are now becoming available. Sample sizes were not large enough to examine influenza viral strains according to comorbidities and other risk factors including pregnancy, but this should become possible as the GIHSN grows and sample sizes increase.
Although the patients included in this study were not necessarily representative of the overall population (i.e. subset of hospitalized patients with a diagnosis possibly associated with influenza), vaccination rates in these patients were lower than recommended by the WHO for European countries [24]. The French patients in this study, regardless their RT-PCR result, had the highest vaccination rate at 47.0%, which was similar to the overall rate for France reported in 2010-2011 (50.4%) [25]. However, it is not unusual to observe some influenza cases in vaccinated individuals because seasonal influenza vaccines have only a moderate protecting effect, especially in elderly adults [26]. The lowest vaccination rates for these patients were in the Russian Federation, where fewer than 2% reported being vaccinated for seasonal influenza. This may have been due to the peculiarity of the groups enrolled (e.g. mostly pregnant women in Moscow and mostly children in St. Petersburg), poor uptake, the ability of the vaccine to prevent hospitalization, or a combination of these factors, although these influences remain to be assessed over the coming years.
The findings of this study are strengthened by the active surveillance methodology, specific definitions at each site of admission possibly related to an influenza infection, confirmation of influenza cases by RT-PCR, and the consistent body of evidence generated across coordinating sites. In addition, all RT-PCR was carried out by WHO National Influenza Centres (France, Moscow, St. Petersburg, and Turkey) or using the WHO protocol (Spain).
The results should be generally applicable because of the diversity of participating hospitals and healthcare settings. Nevertheless, the adaptation of case identification to the specific local setting, mainly driven by practical considerations, might have increased the sensitivity of the results to geographical variation. For future years, to help reduce heterogeneity between sites, we plan a common standard operating procedure and meetings at the beginning of the study to share pilot data and harmonise criteria and follow-up activities.

Conclusion
These first-year findings demonstrate that the GIHSN produces relevant, important, and timely data. Despite the size of the network, results were available before the start of the following influenza season. These results will help understand and prepare for seasonal influenza epidemics. In particular, this platform provides annual data on the severe end of the influenza infection spectrum, as represented by hospitalized cases, for a wide range of populations. More formal burden and risk assessment will become available as documentation of the catchment population of these hospitals and wards improves. By adding the ability to evaluate the impact of vaccination programs, these data support those collected by SARI surveillance systems, although the GIHSN employs a broader definition of ILI that encompasses systemic symptoms beyond fever (i.e., malaise, headache, and myalgia).
The breadth of results provided by the GIHSN are particularly useful for assessing the effectiveness and costeffectiveness of intervention programs, and they open the door to further research and public health activities, especially on risk factors and the impact of influenza and influenza vaccination programs in different populations. The approach used by the GIHSN also allows additional components to be added to respond to other scientific questions. During this first year of the GIHSN, only 21 hospitals and four countries were included, but additional hospitals and countries, including some from