Modelling the influenza disease burden in people aged 50–64 and ≥65 years in Australia

Abstract Background Estimation of influenza disease burden is necessary to monitor the impact of intervention programmes. This study aims to estimate the attributable fraction of respiratory and circulatory disease due to influenza among Australian adults 50–64 and ≥65 years of age. Methods A semi‐parametric generalised‐additive model was used to estimate annual and average rate of influenza‐attributable hospitalisation and death per 100,000 population under the principal diagnosis of influenza/pneumonia, respiratory, circulatory and myocardial infarction (MI) from 2001 through 2017. Results Over the study period, seasonal influenza accounted for an estimated annual average respiratory hospitalisation rate of 78.9 (95%CI: 76.3, 81.4) and 287.5 (95%CI: 279.8, 295.3) per 100,000 population in adults aged 50–64 and ≥65 years, respectively. The corresponding respiratory mortality rates were 0.9 (95%CI: 0.7, 1.2) and 18.2 (95%CI: 16.9, 19.4) per 100,000 population. The 2017 season had the highest influenza‐attributable respiratory hospitalisations in both age groups, and respiratory complications were estimated approximately 2.5 times higher than the average annual estimate in adults aged ≥65 years in 2017. For mortality, on average, influenza attributed 1,080 circulatory and 361 MI deaths in adults aged ≥65 years per year. Influenza accounted for 1% and 2.8% of total MI deaths in adults aged 50–64 and ≥65 years, respectively. Conclusion Rates of cardiorespiratory morbidity and mortality were high in older adults, whilst the younger age group contributed a lower disease burden. Extension of influenza vaccination programme beyond the targeted population could be an alternative strategy to reduce the burden of influenza.

care facilities and other institutions and people who are at high risk of disease during influenza outbreaks. The attack rate of seasonal influenza varies with age and by type of influenza virus. The highest rate of influenza A/H3N2 strain is seen among children aged <5 years and in adults aged ≥65 years, whereas Influenza B disproportionately affects older children and young adults compared to influenza A. 2,3 Current evidence suggests that influenza infection is a precipitant for ischaemic events and that vaccination and prevention of infections may have an important role in reducing the ischaemic burden. 4 From the Australian Bureau of Statistics, ischaemic heart disease (IHD) is the leading cause of death and responsible for 11% of all deaths in 2019. 5 According to the Australian Institute of Health and Welfare (AIHW), cardiovascular disease (CVD) accounts for about 41,000 deaths as the underlying cause of death in Australia in 2018, equating to 26% of all deaths. 6 Moreover, CVD was responsible for about 1.2 million hospitalisations in 2017-2018 and accounts for 11% of all hospitalisations in Australia. Hospitalisation rates of acute myocardial infarction, cardiovascular illness and all-cause death increase during influenza season. [7][8][9][10] In a study, 12.4% of people admitted with AMI had undiagnosed, unrecognised influenza infection, compared to 6.7% of controls who did not have AMI, where it suggests a role of influenza. 11 A study of over 80,000 influenza cases found that 12% had an acute cardiovascular event complicating their admission. 12 Vaccination is the most effective public health intervention for prevention of influenza infection. 13 In Australia, conventional and enhanced influenza vaccines are provided under the NIP. Quadrivalent influenza vaccine (QIV) was available for use from 2016-2017 in Australia, and high-dose (HD) and adjuvanted TIVs were introduced for adults aged ≥65 years under the NIP in 2018, whilst other age groups received standard QIV. Increased efficacy of influenza vaccines has been shown to be associated with reduction of cardio-respiratory illness and all-cause hospitalisation in older adults. 14,15 Despite compelling evidence, the role of infection in ischaemic events is rarely counted in burden of disease estimates and in economic evaluations of the influenza vaccine. The aim of this study was to estimate the burden of respiratory and circulatory diseases attributable to influenza in adults aged 50-64 and ≥65 years in Australia.

| METHODS
We used descriptive epidemiology and statistical modelling method to estimate the age-specific rate of influenza-attributable hospitalisations and deaths due to influenza/pneumonia, respiratory, circulatory and myocardial infarction (MI) in two age groups: 50-64 and ≥65 years, from July 2001 through December 2017. Our study aimed to highlight the gap in the vaccination programme between the two age groups to reduce influenza burden in Australia. We extracted surveillance data on weekly counts of laboratory-confirmed influenza notifications in Australia between 2001 and 2018 from the National Notifiable Diseases Surveillance System (NNDSS), the Australian Government, Department of Health. 16  Population data were sourced from the ABS. 19 Rate per 100,000 population was calculated for 50-64 and ≥65 years.

| Model
A semi-parametric generalised-additive model (GAM) was used to model the burden of influenza in 50-64 and ≥65 years. 20,21 Separate models were used for each age group and principal diagnosis category and estimated the age-specific rate of influenza-attributable hospitalisations and deaths in the study. Time series (weekly counts) of all-age influenza notifications were entered into the model as predictor variables. Furthermore, an interaction term was used as value 0 or 1 between year and influenza notifications to include notifications as separate variables for each year. This at least allowed for changes in laboratory testing and reporting practices by healthcare providers over time during the study period. We adjusted for national holiday effects as reduced trend in hospital admissions was seen during holidays in the observed data. We applied separate indicator variables for each holiday (value 0 or 1) representing a week with a public holiday in it across the years. These variables provided varied effect across the models. We used a natural cubic smoothing spline of the consecutive week number to account for unmeasured non-influenzaattributable hospitalisations or deaths due to seasonal and other factors in the model. The benefit of using GAM is that as it allows the analysis of non-linear association between the dependent and independent variables. 20 The model equation was as follows:   Figure S5.

| Estimated hospitalisations attributable to influenza
The estimated hospitalisation rates varied across the seasons (Table S1) (Table 3).
Yearly estimated influenza-attributable mortality rates are presented in Table S2. In adults aged ≥65 years, respiratory mortality rates  (Table S2).

F I G U R E 1 Laboratory-confirmed influenza notifications in Australia, 2001-2017
In adults aged 50-64 years, influenza-attributable MI and circulatory mortality rates were not statistically significant almost all years in the study except for circulatory, where influenzaattributable mortality rate of 3.9 (95%CI: 1.9, 5.9) per 100,000 (Table S2) was found in 2012. However, in adults ≥65 years, statistically significant association between influenza and MI and circulatory deaths was found (Figure 3). The highest rates in adults ≥65 years were influenza-attributable mortality rate of 31  This study has some limitations. First, in the study, weekly counts of total influenza were applied in the model; thus, type-specific disease burden cannot be estimated in these age groups. Second, the seasonality of influenza infection may coincide with seasonality of respiratory syncytial viruses (RSV), and our study could not be able to adjust presence of RSV in the model; thus, study findings may have been confounded by RSV. Finally, in the study, we used national data on weekly counts of all-age influenza infection for individual year in the model, to account for variation in testing or reporting practices by general practitioners as well as variation across states and territories in Australia over time. Strengths of the study include the use of weekly counts of total influenza infection as a proxy measure or covariate in the model, thus reflecting the incidence of infection in the population to predict better disease estimates across the studied years. Also, 17 years of influenza notifications data in the study could provide a stronger statistical power in model outputs compared to a short-term data analysis.
In conclusion, our study demonstrates that influenza attributed to increased respiratory morbidity in both age groups: 50-64 and ≥65 years and high circulatory mortality among the older population.
Influenza and its associated complications can be prevented by seasonal vaccines. Improved vaccine uptake in the recommended age group, as well as in adults 50-64 years of age including the high-risk population, might be the best available intervention to reduce cardiorespiratory disease burden associated with influenza.