Historically, influenza surveillance has focused on outpatient visits for influenza-like illness (ILI; defined as temperature ≥38°C with either cough or sore throat and no alternative diagnosis) and, among those with ILI, the proportion with laboratory-confirmed influenza, stratified by viral type/subtype.1 Influenza surveillance in outpatients has informed current understanding about the seasonality and characteristics of influenza viruses, data that have helped determine the timing and composition of annual influenza vaccination.2 Recently, the World Health Organization (WHO) has recommended that influenza surveillance also include sentinel surveillance for severe acute respiratory infection (SARI), which is often defined as ILI plus one additional symptom or sign of severe illness in a hospitalized patient.3 The rationale is that severe outcomes of influenza have the greatest public health and economic impact and that identifying risk factors for severe illness informs vaccine policy.

The burden of influenza, specifically hospitalization rates attributable to laboratory-confirmed influenza, is well documented in industrialized countries4–8 but less so in developing countries and in subtropical regions.9–13 China is a lower middle-income country with the largest population (1·3 billion) in the world. The burden of laboratory-confirmed influenza among a defined general Chinese population has not been thoroughly studied. Although seasonal influenza vaccine use has increased dramatically, the vast majority of the Chinese population is not vaccinated annually.14

During 2010, in the post-2009 H1N1 pandemic period, we initiated active surveillance for SARI in a central Chinese city. We analyzed data from this system during a 24-month period to characterize the epidemiology and estimate hospitalization rates attributable to laboratory-confirmed influenza.

Our study is the first in China to estimate hospitalization rates attributable to influenza by viral type/subtype among a well-defined population. We found that a substantial number of patients are hospitalized annually with influenza, especially children aged <5 years. Hospitalization rates varied by influenza type/subtype, with much higher rates associated with influenza B than A(H3N2) during 2011–2012. Influenza hospitalizations varied throughout the year with peaks in summer, winter, and spring months, in contrast with the distinct winter-only peak observed in Northern China.18

Among all cause hospitalizations in young children aged <5 years in the surveillance hospitals, we estimate that 35% were for SARI during the two-year-study period, including 4% with laboratory-confirmed influenza. Influenza-associated SARI hospitalization rates were high for children aged <5 years in each of the two seasons, but were sharply lower for those aged ≥5 years. Similarly high hospitalization rates have also been documented for seasonal influenza in subtropical regions such as Hong Kong SAR, Bangladesh and Mexico, with estimated rates varying from 380 to 670 per 100 000 population among children aged <5 years.11–13 Studies conducted in the temperate USA, which used different methodologies and denominators, reported estimated hospitalization rates for laboratory-confirmed influenza of 8·6 per 10 000 children, 3–4 per 1000 children aged <2 years/year and 5–151 per 100 000 per person-months for children aged <5 years.5,6,8 Our findings are consistent with a recent global systematic review, which estimated 3500 influenza-associated acute lower respiratory infections (ALRI) occurred in developing countries and 200 influenza-associated severe ALRI cases in the WHO Western Pacific region per 100 000 for children younger than 5 years in 2008, respectively.19 Studies in temperate countries, such as USA, have reported the highest influenza-associated hospitalization rates in infants aged <6 months,4 with high rates also observed in persons aged ≥65 years.20 We found that, in central China, influenza-associated hospitalization rates among persons ≥65 years were much lower than for young children, but are slightly higher than that the rates estimated in subtropical Bangladesh 12 and in the temperate USA.20 The highest influenza-associated hospitalization rate in our study was in infants aged 6–11 months and the second highest in those <6 months. This is different from that observed in the USA, where the highest rate was observed in infants<6 months, although that study did not provide specific data on children aged 6–11 months.4

To our knowledge, our study is one of the first to report hospitalization rates attributable to influenza by viral type/subtype. A few small cohort studies reported the proportion of children presenting at the emergency room in France21 and in USA,22 and the proportion of health care visits in the USA,6,23 caused by influenza B and A(H3N2). Influenza B-associated hospitalization rates in our study during 2011–2012 were more than twice the rates associated with A(H3N2). This finding suggests that illness caused by influenza B virus were at least as severe as A(H3N2) viral infections in one season, which is different than one study in which estimated excess hospitalization rates associated with influenza B were lower than for A(H3N2) using a time series modeling approach.24

Influenza viral strains associated with SARI in Jingzhou were generally similar to the predominant strains detected by the national influenza surveillance system in the region. In our study, similar to A(H3N2) viruses, influenza B viruses were associated with a disproportionate number of hospitalizations in children aged <5 years during the 2-year surveillance period, which is different from the perception that influenza B illness appears to be highest among older children and young adults. Multiple studies have reported that children with influenza B are older than those with influenza A.25–28 In two studies that described influenza illness among children and adults, the proportion of illnesses caused by influenza B relative to influenza A was highest for individuals aged 5–29 years 29 and 2–39 years,30 respectively.

There are a number of reasons why influenza-associated hospitalization rates may vary across time, countries, and climates. Direct comparisons may be limited by differences in case definitions, diagnostic methods, age stratification, study period, influenza activity and seasonality, influenza vaccine coverage, health care utilization, structure of the health care system, population age structure, and clinical criteria for hospital admission. Host immunity and virulence of influenza viral strains, such as pH1N1 virus or antigenically drifted strains, may also contribute to differences. Compared with other studies, we used a broad SARI case definition for enrollment. Compared to estimates of excess influenza-associated deaths using an indirect method approach – statistical models,18 few studies have used a direct method approach to estimate mortality attributable to laboratory-confirmed influenza. This may be because deaths caused by influenza may be unrecognized or attributed to co-morbidities or to secondary complications, particularly among the elderly.31,32 Our study population had a lower prevalence of known chronic medical conditions compared with that reported in the USA4–7 and may be partially explained by an overall lower prevalence of chronic diseases among the general Chinese population.33

Obtaining accurate data about hospitalizations associated with influenza is important for allocating resources for diagnosis, treatment, and the prevention of respiratory pathogens. Further studies are needed to elucidate how much variability in hospitalization rates is attributable to health-seeking behavior and physician practice and how much to differences in host and pathogen factors. One important next step will be identifying what pathogens are responsible for the 87% of SARI patients who tested negative for influenza viruses, including Streptococcus pneumoniae and Haemophilus influenzae type B, the leading bacterial causes of childhood pneumonia.34,35

The results in our study have important implications for influenza prevention in China. Seasonal influenza vaccination is not included in the national immunization program, and individuals must purchase it themselves. China CDC currently recommends annual influenza vaccination for persons with chronic illness, pregnant women, individuals aged <5 or ≥60 years old, health care workers, and close contacts of high-risk individuals,36 an estimated population of 570 million.14 China does not currently have the capacity to produce this much influenza vaccine.14 Many in China perceive that influenza does not contribute substantially to morbidity, but our study demonstrates that many persons are hospitalized annually for influenza, particularly children aged <5 years. Our results support China CDC's recommendations and strongly suggest that influenza vaccination should be targeted at young children if the goal is to reduce severe complications of influenza.

Our study is subject to several limitations. First, this study was only conducted for a 24-month period, and the influenza disease burden findings may not reflect hospitalization rates over many years. Second, we may have under-estimated influenza-associated hospitalizations, because we required that patients have an elevated temperature when screened within 24 hours after admission. A substantial population of hospitalized influenza patients do not have fever, particularly the elderly,37,38 and influenza viral infection can exacerbate underlying conditions resulting in hospitalization and death from non-respiratory illness, such as myocardial infarction and diabetic ketoacidosis.39–42 Diagnostic techniques, such as reverse transcription polymerase chain reaction (RT-PCR) and viral isolation, may yield false-negative results if samples are collected after viral shedding has ceased. This could also lead to under-estimation of influenza-associated hospitalizations. Third, SARI surveillance was not conducted at all hospitals, in the city. Fourth, the low median age and high estimated influenza-associated hospitalization rates for young children might be partially explained by the predominance of 2009 H1N1 and influenza B viruses during five of the seven periods of influenza activity under surveillance for this study. However, additional years of surveillance, especially during periods when H3N2 viruses are predominating, and health utilization survey data, are needed to assess the impact of influenza complications among elderly persons in Jingzhou and other areas of China. Finally, the findings from this study should not be generalized to all of China, where climate and other factors may cause variation in influenza burden.

Given the enormous population, geographic area, and diversity of climates, multiyear population-based surveillance with laboratory confirmation of viral and bacterial respiratory pathogens should be implemented in different geographical and climactic regions of China. Data from such studies will provide evidence to help prevent influenza and other respiratory infections in China. Continuing surveillance over several years and conducting similar studies in temperate and subtropical regions in China will help confirm our findings on influenza seasonality and may help improve our understanding of factors driving seasonality.