Impact of influenza on hospitalization rates in children with a range of chronic lung diseases

Background Data on burden of severe influenza in children with a range of chronic lung diseases (CLDs) remain limited. Method We performed a cohort study to estimate burden of influenza‐associated hospitalization in children with CLDs using population‐based linked data. The cohort comprised all children in New South Wales, Australia, born between 2001 and 2010 and was divided into five groups, children with: (a) severe asthma; (b) bronchopulmonary dysplasia (BPD); (c) cystic fibrosis (CF); (d) other congenital/chronic lung conditions; and (e) children without CLDs. Incidence rates and rate ratios for influenza‐associated hospitalization were calculated for 2001‐2011. Average cost/episode of hospitalization was estimated using public hospital cost weights. Results Our cohort comprised 888 157 children; 11 058 (1.2%) had one of the CLDs. The adjusted incidence/1000 child‐years of influenza‐associated hospitalization in children with CLDs was 3.9 (95% CI: 2.6‐5.2) and 0.7 (95% CI: 0.5‐0.9) for children without. The rate ratio was 5.4 in children with CLDs compared to children without. The adjusted incidence/1000 child‐years (95% CI) in children with severe asthma was 1.1 (0.6‐1.6), with BPD was 6.0 (3.7‐8.3), with CF was 7.4 (2.6‐12.1), and with other congenital/chronic lung conditions was 6.9 (4.9‐8.9). The cost/episode (95% CI) of influenza‐associated hospitalization was AUD 19 704 (95% CI: 11 715‐27 693) for children with CLDs compared to 4557 (95% CI: 4129‐4984) for children without. Discussion This large population‐based study suggests a significant healthcare burden associated with influenza in children with a range of CLDs.


| INTRODUC TI ON
Globally, influenza is a major public health problem. Every year during seasonal epidemics, 3-5 million people develop severe influenza requiring hospitalization. 1 Young children, especially those aged <5 years, are particularly vulnerable to infection; around 10%-30% of children are infected with the virus during each influenza season. 2 Children also play an important role in transmission by shedding large quantities of virus for 7-10 days and introducing the virus into households. 3 Influenza virus infection may be especially problematic in children with chronic lung diseases (CLDs). It can exacerbate the respiratory symptoms of the underlying lung condition resulting in unscheduled medical presentations and contributing to the total burden of CLDs on the health system. Children with CLDs such as asthma have been reported to be at double the risk of hospitalization with influenza compared to children with other chronic conditions. 4 Data on the burden of influenza among children with other CLDs are more limited, and largely limited to the burden in children with cystic fibrosis (CF). 5 Such data are crucial for informing policy makers, clinicians, and public health professionals of the magnitude of the problem and for monitoring burden of disease over time and evaluate vaccine effectiveness thus influencing influenza vaccine policy. We therefore conducted a retrospective population-based cohort study designed to measure the incidence rates and direct healthcare cost related to influenza-associated hospitalization in children with CLDs.

| Study design
The retrospective cohort study used linked population-based administrative data sets. In NSW, The Centre for Health Record Linkage (CHeReL) (www.cherel.org.au) conducts linkage of administrative health data sets for research purposes and provides each child with a unique Patient Project Number (PPN). The de-identified data sets with the unique PPNs were provided to the study investigators. Records of the same child in the different data sets were combined using the unique identifier key. 6 The NSW Perinatal Data Collection (PDC) which records all births in NSW was used as the primary data source for identifying the study cohort. Information relating to maternal and child factors such as maternal smoking during pregnancy, number of previous pregnancies, Indigenous status of the mother, postcode of area of residence, birth weight, gestational age at birth, and sex of the baby was ascertained from PDC. Data relating to diagnosis of CLDs and influenza-associated hospitalization were ascertained from the Admitted Patient Data Collection (APDC) and Neonatal Intensive Care Unit Data (NICU). The data sets have been described previously. 6

| Exposure assessment
The birth cohort was divided into five mutually exclusive sub-groups.
As diagnosis of asthma is difficult in children aged <2 years 8 and as the national immunization guidelines recommend seasonal influenza vaccine for children with severe asthma (requiring frequent hospitalization), 9 only these children were included in the analysis. Additionally, we excluded ICD codes associated with recurrent wheeze as many of these could have been due to respiratory viral infection and not due to asthma.
2. Children with bronchopulmonary dysplasia (BPD): Children who were born at gestational age (GA) ≤32 weeks and required oxygen and/or any type of respiratory support at 36 weeks' GA 10 or had any history of hospitalization with ICD-10 diagnostic code associated with bronchopulmonary dysplasia originating in the perinatal period (P27.1). The NICU data set and the APDC data sets were used to identify children with BPD. 5. All other children without chronic lung diseases.

| Primary outcome
Influenza-associated hospitalization: The ICD-10, primary diagnostic codes associated with influenza-associated hospitalizations were used to identify all influenza-associated hospitalizations from APDC. May-September in Australia) were also considered to be associated with influenza.

| Other co-variates
Data on maternal and child factors considered to be independent risk factors for acute lower respiratory infections 11 including multiparity of the mother (previous pregnancy lasting >20 weeks), maternal smoking during pregnancy, Indigenous status of the mother, residential postcode of the mother at birth, small for gestational age, and sex of the cohort child were ascertained from the PDC. Frequency of allcause-associated hospitalizations in the cohort child was retrieved from the APDC. Socioeconomic disadvantage was inferred from the maternal postcode using the Socioeconomic Index of Areas (SEIFA) and Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD) compiled by the Australian Bureau of Statistics. 12

| Bias
This was a large population-based cohort study with minimum selection bias.

| Study size
This was a whole-of-population study including all children born in NSW between 2001 and 2010.

| Analyses
The overall and year-specific incidence rates for influenza-associated hospitalizations were calculated for years 2001-2011. We calculated incidence as the number of new influenza-associated hospitalizations divided by the child-years at risk in each of the disease groups.
A lag of 14 days between two successive influenza-associated hospitalizations was considered to be a new hospitalization. We used Poisson estimation to calculate incidences and incidence rate ratios (including 95% confidence interval around the estimates) of influenzaassociated hospitalizations. The incidence rates and rate ratios were adjusted for parity of the mother, maternal smoking during pregnancy, Indigenous status, IRSAD, sex of the child, small for gestational age, and frequency of previous hospitalizations in the cohort child. There were 0.5% missing data for variables including Indigenous status of the mother, socioeconomic disadvantage of the area of residence, and maternal smoking during pregnancy. Observations with one or more variables missing were dropped from the analyses.    (Table 1). The unadjusted incidence/1000 child-years of influenzaassociated hospitalization in each CLD group was 1.1 (0.7-1.4) for children with severe asthma, 5.4 (3.6-7.2) for BPD, 6.9 (2.5-11.4) for CF, and 7.7 (5.9-9.4) for children with other congenital and chronic lung conditions. The adjusted age-specific incidence rates and rate ratios are presented in Table 2.

| Inpatient resource consumption of influenza illness
The length of stay (days) for each episode of influenza-associated hospitalization in children with and without CLDs is presented in

| D ISCUSS I ON
This large population-based study has demonstrated that children with CLDs are at least five times more likely than children without CLDs to be hospitalized with influenza. A previous study from the USA has also demonstrated that children with acute cardiopulmonary diseases were 2-4 times more likely to be hospitalized with influenza-associated illness than other children. 4 The direct medical cost of AUD 19 704/episode of influenza-associated hospitalization in children with CLDs was also four times higher compared to children without, representing a high economic burden on the healthcare system. Furthermore, the hospitalization rates in our cohort were highest in the youngest children and decreased with age, a finding that is consistent with the USA study. 4 Younger children especially those aged <2 years are at higher risk than older children of severe influenza, 14 which is increased in the presence of an underlying CLD. Vaccination against influenza is the primary strategy to control seasonal outbreaks. Influenza vaccination has been proven to be a safe and effective in people with CLDs. 15 The Australian  immunization guidelines emphasize the need for seasonal influenza vaccine for people aged ≥6 months with chronic medical conditions. 9 Maternal immunization to protect infants in their first 6 months of The rate of influenza-associated hospitalization in each of the groups of children with chronic lung diseases was significantly higher (P < 0.05) compared to rate in all other children without chronic lung diseases.
risk of being hospitalized with respiratory illness compared to other children. [25][26][27] While several studies have shown the significant burden of respiratory syncytial virus among children with BPD, 6, 28 the impact of influenza illness remains less clearly defined.
Our study also showed that the rate of influenza-associated hospitalization was 10 times higher in children with other congenital and chronic lung diseases than all children without CLDs, yet there are almost no available data on the burden of influenza in these children.
Our study has several important limitations. There are no national disease registries for children with CLDs (with the exception of CF) so our cohort was constructed from a comprehensive hospitalization dataset based on ICD codes. Ascertainment of influenza-associated hospitalization was also based on ICD codes.
Even though not all influenza-associated hospitalizations are laboratory confirmed but as influenza is a notifiable condition in NSW, it is likely that most hospitalizations receiving an influenza-specific code were laboratory confirmed; however, we were not able to substantiate this. Not all children hospitalized with acute lower respiratory illness are routinely tested for influenza so it is likely that the true burden of influenza-associated hospitalization in children with CLDs is even higher than our estimates. We considered all influenza-coded hospitalizations where an influenza virus was not identified to be associated with influenza which may have inflated our estimates; however, none of the influenza-associated hospitalization in our cohort was associated with an ICD code for influenza where an influenza virus was not identified. We did not have access to data on primary care or emergency presentations and therefore we could not assess the rate of ambulatory influenza infection in children with CLDs, where the burden is likely to be higher. Although we adjusted our estimated rates for a number of known potential confounders, we did not have access to direct information on other potential confounders such as household exposure to tobacco smoke or presence of siblings at home.
Instead, we used maternal smoking during pregnancy as a proxy for household smoke exposure and previous pregnancy/parity (lasting at least 20 weeks) as a proxy for having older siblings at home. We used the IRSAD score of the mother's postcode at the time of delivery, a measure of average socioeconomic disadvantage in that postcode, as a proxy for family level socioeconomic disadvantage. 29 We estimated incidence rate ratio for influenzaassociated hospitalization in children with CLDs compared to all other children which included children with other chronic conditions and may have led to lower rate ratios. Influenza vaccination status is not routinely recorded in the national immunization data base, and we did not have access to antiviral use in these children.
In summary, although only around 1.4% of the total pediatric population in our cohort had one of the chronic lung diseases, these children form a special group because of their ongoing need for healthcare services. Our data clearly show that influenza illness in these children add to the existing burden of chronic diseases on the health system. While there are no effective vaccines or antivirals for most childhood respiratory viral infections, fortunately we do have a vaccine against influenza. There are also effective antivirals which, if administered in a timely manner, may reduce disease severity. [30][31][32] However, such high burden of hospitalizations indicates that current efforts at influenza prevention are inadequate among children with CLDs. Further studies investigating the effectiveness of newer vaccines, treatment, or chemoprophylaxis in these children will help lower the burden of disease.

ACK N OWLED G EM ENTS
The authors would like to thank Sydney Children's Hospital Foundation and Ministry of Health, NSW. We also thank the babies and their families, the nursing and midwifery, obstetric and medical records staff of the obstetric and children's hospitals in NSW and the ACT.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest relevant to this article to disclose.