Pneumonia and influenza hospitalizations among children under 5 years of age in Suzhou, China, 2005–2011

Background The disease burden of influenza among children in China has not been well described. Objective To estimate the influenza‐associated excess hospitalization rate and compare the hospitalization length and costs between pneumonia and influenza (P&I) and other community‐acquired diseases (CAD) in Suzhou, China. Methods We retrospectively collected hospital discharge data on pediatric patients' discharge diagnosis, hospital costs, and length of hospital stay in Suzhou. P&I hospitalization was defined as a primary discharge diagnosis of pneumonia and influenza disease (ICD‐10 codes J09–J18). Other CAD were common community‐acquired diseases among children. Negative binomial regression models were used to estimate the weekly P&I hospitalizations in Suzhou. Excess P&I hospitalizations due to influenza were calculated as the difference in P&I hospitalizations between the epidemic period and the baseline period. Baseline was defined as when the influenza‐positive rates were <5% for two consecutive weeks. Results From October 2005 to September 2011, we identified a total of 180 091 all‐cause hospitalizations among children <5 years of age in Suzhou City. The rates of P&I and influenza‐associated excess hospitalizations were highest in the 2009–2010 pandemic and 2010–2011 post‐pandemic seasons. Infants <6 months of age had the highest P&I hospitalization rates, the longest hospital stays (7.5–8.0 days), and the highest hospitalization costs for P&I. Compared with other CAD, children admitted for P&I had longer hospital stays and higher hospitalization costs. Conclusions The influenza‐associated P&I hospitalization rates and economic burden were high among children. Targeted influenza prevention and control strategies for young children in Suzhou may reduce the influenza‐associated hospitalizations in this age group.


| INTRODUCTION
Influenza is responsible for substantial morbidity and mortality worldwide every year. Estimating influenza-associated mortality is important for understanding the epidemiology of influenza. However, influenzarelated pediatric deaths are not common, 1,2 although the substantial number of influenza-associated hospitalizations among children each year leads to significant burden and economic cost to their families. [3][4][5][6][7][8][9][10][11][12] Previous prospective studies have estimated high rates of influenza-associated hospitalization among children from China and other countries. 4,10,11,13 However, the influenza epidemic varies annually, and such studies are resource intensive. Additionally, when specimens are not collected properly (such as when they are not collected at the right time period of disease course or at the right sites), studies will underestimate the hospitalization rates associated with influenza. Thus, many researchers use statistical models based on morbidity from pneumonia and influenza (P&I) hospitalizations to estimate the number of hospitalizations related to influenza. 2,9,14,15 Pneumonia and influenza are clinically diagnosed respiratory infectious diseases. Although influenza is not routinely diagnosed in hospitals, it is a vaccine-preventable disease. Some studies have shown that P&Irelated diagnoses can be used to indirectly estimate the influenza virus activity and the disease burden caused by influenza virus. 9,16 With the development of electronic health information systems, more researchers are using P&I discharge diagnoses to identify the influenza-related hospitalizations. Furthermore, identifying P&I hospitalizations using the international code of disease allows for comparison with other studies.
Pneumonia and influenza hospitalizations can impose the substantial socioeconomic burden on families and society. It is estimated that in the United States, the direct costs related to hospitalization due to influenza among children less than 5 years average $809.1 million annually. 17 In addition to the direct costs of medical care, influenza in this age group has also significant indirect costs, especially in terms of parental work loss while caring for the sick child. 18 However, in China, data on the influenza-associated P&I hospitalization rates, especially regarding the economic burden of influenza-related P&I hospitalization among children, are scarce. 2,19 The limited data on the influenza disease burden present an obstacle for developing influenza control policies.
To address this gap, we collected pediatric discharge data from hospitals, employed a negative binomial regression model to estimate the influenza-associated P&I hospitalizations among children under 5 years of age between 2005 and 2011 in Suzhou. To better understand the characteristics and economic burden of P&I hospitalization, we compared the hospitalization stays and costs of P&I hospitalizations with those of other common community-acquired diseases among children.

| Study site
This study was conducted in Suzhou, a major city located in the southeast area of Jiangsu Province in eastern China. Based on 2010 estimates in Suzhou, the population of children <5 years old comprised 209 276 Suzhou residents and 256 719 migrants (data from the immunization program database). Suzhou consists of five municipal districts and 5 county-level cities. In 2011, there were 181 hos pitals that admitted inpatients, and 36 had over 100 pediatric inpatients in Suzhou city.

| Data collection
We selected at least one hospital that admitted pediatric patients from each county and from the municipal districts to participate in this study ( Figure S1). The selected hospitals were among the largest hospitals in the county and the districts and admitted at least 2000 pediatric cases per year. We retrospectively collected discharge records for all children less than 5 years old who were admitted to the selected hospitals from 2005 to 2011. We exported the following variables from the electronic health information system (HIS) database of each hospital: admission number, admission date, discharge date, gender, date of birth, discharge diagnosis and/or ICD-10 codes, and total hospitalization costs (including nursing, physician services, room cost and supplies, diagnostics, and therapeutics). it accounted for at least 50% of the respiratory specimens that were typed. According to the virological surveillance data, we defined the beginning of the influenza season as October 1.

| Ethics statement
The study was approved by the Institute Review Board (IRB) of the School of Public Health, Fudan University. As this was a retrospective, electronic medical data abstraction study that did not collect any personal identifiers or involve any patient contact, this study was exempt from obtaining informed consent.

| Statistical analysis
Data analyses were performed using SAS software version 9.0 (SAS Institute Inc., North Carolina, US). Descriptive statistics were used to summarize the continuous variables and discrete variables. Categorical variables were presented as numbers or percentages. Continuous variables were presented as the mean and standard deviation (S.D.), the mean and 95% confidence interval (CI), or the median with interquartile range (IQR). Chi-square test was used to compare the population characteristics and the P&I hospitalization rates between groups.
Kruskal-Wallis test and Mann-Whitney test were used to analyze the length of hospital stay and the hospitalization costs.

| Estimating influenza-associated P&I hospitalization rates
We used negative binomial regression model to estimate the weekly counts of P&I hospitalizations. 2,19,21 The negative binomial regression models we used were as follows: where Yt i represents the t i week counts of P&I hospitalizations from 2005 to 2011 for a specific age group. The term "α" is the population offset (α = logpop). The term "t i " is the number of weeks in a series from the beginning of October 2005 through the end of September 2011. We estimated the following β coefficients: β 0 was the intercept, β 1 accounted for seasonal changes in P&I hospitalizations, β 2 accounted for nonlinear time trends, β 3 , β 4 , β 5 , and β 6 accounted for seasonal changes in P&I hospitalizations (full year and half year), and β 7 , β 8 , β 9 , and β 10 were coefficients associated with the percentages of specimens testing positive for specific influenza viruses in a given week. The goodness of model fit was tested by R 2 (1-SS reg /SS tot ). R 2 was above 95% for each age group of children.
The P&I hospitalization rates were calculated as the cumulative counts of P&I hospitalizations for each season divided by the corresponding population derived from the immunization database.

| RESULTS
In total, we selected nine hospitals for inclusion in this study ( Figure S1).

| Influenza-associated P&I hospitalizations
The estimated number of influenza-related P&I hospitalizations among Older children had lower P&I hospitalization rates (P < .01). The P&I hospitalization rates among children 2 < 5 years of age were only a tenth to one-third of the rates among infants <6 months (Table 2).

| Influenza excess P&I hospitalizations
According to the influenza virological data, there were two influenza  (Table S1). The annual excess rates of P&I hospitalizations due to influenza among those younger than 5 years old were highest in the 2009-2010 pandemic influenza season (9.6 per 1000 children).

| Length of hospital stay
The average length of hospitalization due to P&I was 7.1 days during the study period, which was longer than that of other CAD (5.7 days) (P < .01) (

| The hospitalization costs
The median P&I hospitalization cost was 2625.0 Renminbi, the official currency of China (RMB) during our study period, which was higher than that of other CAD hospitalizations (2040.5 RMB) (P < .001) (

| DISCUSSION
Our findings demonstrate that influenza activity is associated with P&I-associated hospitalization rates among children less than 5 years    Compared with children aged 2 < 5 years old, the children aged 6 < 24 months had higher P&I hospitalization rates and costs. These results were consistent with the findings from a study in central China of the same season. 13 The influenza vaccine is the most effective method for influenza control and prevention. Although the effectiveness of the vaccine varies annually, the average influenza vaccine effectiveness has been estimated to be approximately 50%-72%. 26,27 The data on influenza vaccine effectiveness in China are consistent with those of developed countries. [28][29][30] However, the influenza vaccination coverage among children ≤5 years of age in China was low, ranging from a low of 8.6% to a high of 26.4% between the 2009 and 2012 influenza seasons, with no increasing trends by year. 28,31 Therefore, children less than 5 years of age, especially children under 24 months, could benefit from increasing rates of influenza vaccination.
This study has several limitations. First, P&I hospitalizations are associated with numerous pathogens in addition to influenza, such as RSV, parainfluenza virus, and adenovirus. However, we did not conduct laboratory testing for respiratory pathogens in our study, and therefore, we cannot specifically estimate the influenza-associated hospitalizations. Second, the influenza virus surveillance data originated from the surrounding provinces and not specifically Suzhou city, which may suggest that the data did not directly reflect the influenza epidemic status in Suzhou and could cause some bias in estimating the excess P&I hospitalizations among children in Suzhou. Third, although we considered seasonal changes that could represent potential confounding factors, there may still have been some unmeasured factors such as income or distance between residence and hospitals that could have affected the estimates of influenza-associated hospitalizations for P&I.
Pneumonia and influenza hospitalizations appear to be an increased burden on young children and their families in Suzhou.
Children less than 5 years of age, especially children under 24 months, could benefit from increasing the influenza vaccination coverage.