Hospital surveillance of influenza strains: a concordant image of viruses identified by the Swiss Sentinel system?

Background The Swiss Sentinel system for influenza virus surveillance reports influenza‐like illness in the community through a network of primary care practitioners, but the epidemiologic, demographic, and virological characterization may differ from that observed in hospitalized patients with influenza. Objective To compare demographic and virological data from hospital influenza cases with Sentinel system data during the 2014–2015 season. Methods We included 2623 in‐ and outpatients with a screening request for influenza A/B in a university teaching hospital in Geneva, Switzerland, and 933 participants from the Swiss Sentinel surveillance system and compared the demographic and virological data of the two populations, including the respective distribution of influenza subtypes, and conducted a phylogenetic comparison at the HA1 level of influenza viruses recovered in community and hospital cases. Results There were similar proportions of influenza strains recovered in the hospital and in the community (H3N2, 57.1% and 56.9%; H1N1pdm09, 15.5% and 14.2%; B, 27.4% and 28.8%, respectively). HA1 sequence analysis confirmed that all three strains were genetically similar between the two populations. During this particular season, influenza cases were detected earlier in the hospital than in the Sentinel system. Conclusions Although an influenza surveillance system based on the community can predict the type of influenza strains that will be associated with hospitalizations, it fails to estimate the potential virulence of circulating strains. Further, the population characteristics in the community differ from those in hospitalized patients. This suggests that any national influenza surveillance system should include both community‐ and hospital‐based surveys.


| INTRODUCTION
Influenza virus infections are a major burden worldwide in terms of human morbidity, mortality, and public health costs. 1,2 In Switzerland, a network of primary care medical practitioners (Sentinel surveillance system) reports medical consultations for influenza-like illness (ILI) on a weekly basis to the Federal Office of Public Health (FOPH). A subgroup of these practitioners randomly collects respiratory samples from patients diagnosed with ILI for influenza virus detection and This work was performed at the Geneva University Hospitals. characterization at the Swiss National Reference Centre for Influenza (NRCI) in Geneva. In addition, it is compulsory to report influenza A and B infections diagnosed by hospitals and private laboratories to the FOPH. [3][4][5] The ILI-based Swiss Sentinel surveillance of Influenza monitors the evolution of influenza activity and the duration of the influenza season.
However, it cannot identify severe acute respiratory infections due to influenza requiring hospitalization, often observed in individuals with underlying chronic conditions and the elderly. A hospital-based sentinel surveillance would fill this gap, complementing the actual influenza compulsory reporting by hospitals, by providing important data on high-risk groups for influenza infection, for which prevention and treatment should be prioritized. This would permit a better understanding of the clinical features of influenza infection in hospitalized patients with a challenged health status including associated comorbidities and mortality rates and provide a more accurate estimation of the global burden of the disease. 6,7 Characterizing hospital-based influenza strains would also provide the opportunity to assess whether these strains mirror those circulating in the community, increasing the probability to identify more virulent isolates. This strategy is supported by the increasing adherence to international or country-specific, hospital-based surveillance systems by several countries. [6][7][8][9] The aim of this study was to compare demographic, epidemiological, and in particular, virological data from hospital-based influenza cases with data collected by the Swiss Sentinel system during the 2014-2015 influenza season.

| Setting and individuals/samples included
We conducted a retrospective study at the Geneva University Hospitals, Geneva, Switzerland, where inpatients and outpatients are routinely screened for influenza A/B for medical follow-up or diagnostic purposes. Demographic, epidemiological, and virological data from hospital patients were compared to those obtained from the Swiss Sentinel surveillance of influenza. The latter surveillance system relies on the participation of general practitioners, internists, and pediatricians in private practices (n=84 for the 2014-2015 season).
Practitioners are requested to collect naso/oropharyngeal swabs from one of five consulting patients with ILI symptoms from week 40 to week 16 of the following year. The practitioners are recruited according to the Swiss population distribution.
Hospital cases: All in-and outpatients of all ages with at least one respiratory sample (naso/oropharyngeal swab, nasal aspirate, or bron- All influenza cases identified among inpatients were prospectively classified as community-or hospital-acquired (nosocomial). A nosocomial case was considered when the onset of ILI symptoms, confirmed by a positive PCR result, occurred more than 48 hours after hospital admission. Nosocomial cases were analyzed separately and not compared to the Sentinel population.
Sentinel cases: Individuals consulting Sentinel practitioners and with respiratory samples (naso/oropharyngeal swabs) screened at the NRCI during the 2014-2015 influenza season were included in the study. All influenza-positive samples were subtyped, and one of five positive samples were submitted for HA1 sequencing.

| Virological, genetic, and HA1 phylogenetic analysis of influenza-positive samples
Viral genomes of samples selected for subtyping and sequencing were individually processed according to the NRCI procedure.
Two hundred microliters of the initial respiratory specimens was extracted using the NucliSens easyMAG magnetic bead system  (Table S1). Strain-specific HA1 cDNAs were further amplified using either a nested PCR for influenza B/HA1 or a first-round PCR with strain-specific primers, followed by two independent hemi-nested PCRs for influenza A/H1N1pdm09 and A/H3N2 HA1, respectively (Table S1). Amplicons were sequenced with strain-specific primers using conventional Sanger sequencing performed with the ABI 3500xL Genetic Analyzer (Applied Biosystems). 10 Primer sequences and PCR conditions were applied according to the standard operating procedures of the WHO Collaborating Centre at the National Institute for Medical Research (London, UK). HA1 sequences were analyzed with the software platform Geneious 6.1.6 11 and aligned using the mafft v7.017 program 12 . Maximum-likelihood trees were estimated using the PhyML program (1000 bootstrap replicates). 13 Reference sequences used in the phylogenic trees were imported from the GISAID platform (Table S2).

| Data analysis
Differences between groups were tested using the Mann-Whitney U test for continuous variables and the chi-square test with Yates correction or Fisher exact test for categorical variables using Graphpad prism version 6.00 for Windows (GraphPad Software, La Jolla CA, USA). A two-sided P value of <.05 was considered significant. This study was approved by the Research Ethics Committee of Geneva (project # 15-252-2015-00019).   were not hospitalized at the time of respiratory screening (e.g., outpatients consulting in emergency and ambulatory units), (Fig. 4A)

| Influenza subtypes and phylogenetic analysis of Sentinel and hospital samples
One sample out of five was randomly chosen among the positive sam-  (Fig. 1). The

| DISCUSSION
Our results show a higher ratio of positive to negative samples for the Sentinel population compared to hospital patients. This observation could be explained by the fact that the "source" individuals for Sentinel samples were screened on an acute respiratory illness/ILI syndromic basis to specifically optimize the detection of influenza infections. This contrasts with hospital-admitted patients who may present a larger spectrum of diseases and whose samples were collected for the screening of respiratory viruses in general, including influenza. Therefore, the probability to obtain a positive sample from a Sentinel individual was expected to be higher. Of note, acute respiratory illness/ILI-based screening may limit the identification of cases with atypical symptoms either due to host immune status and/or virus strain-specific characteristics. 15 A similar proportion of influenza A and B strains co-circulated in the hospital and Sentinel populations, with a predominance of influenza A/ No significant genetic differences could be observed between influenza strains circulating in hospital and Sentinel populations.
Influenza strain comparison was based on the HA1 gene, one of the main drivers in influenza pathogenicity. Nonetheless, influenza viral particles contain eight segments and the particular constellation of these genes is the major virulence determinant of each strain. 19 Thus, we cannot exclude that significant differences may be found on other influenza genes than HA.
In contrast to the age groups of 0-4, 5-14, 15- Our study has some limitations. As the study was retrospective, some demographic and epidemiological data that would be interesting to compare between both populations were either unavailable or incomplete. Notably, influenza-associated comorbidity information was not reliably reported for the Sentinel population and mortality data were missing. Among the hospital population, children and young adults were underrepresented, while the elderly may possibly be underrepresented in the Sentinel data. Of note, this study was con- hospitals with laboratory confirmation for influenza virus, as proposed by the WHO 25 , but not implemented in Switzerland so far.
We conclude that a hospital-based system for influenza surveillance could be a useful complement to the current Sentinel system.