Inter‐seasonality of influenza in Australia

Background It appears inter‐seasonal influenza notifications have been increasing in summer months in Australia. This study aims to determine changes in inter‐seasonal influenza activity in Australia over time. Methods Routine influenza surveillance data and hospitalisations data were analysed to study the epidemiology of inter‐seasonal influenza and to examine the impact of inter‐seasonal influenza on morbidity in Australia at a national level. To adjust for changes in testing over time, we calculated a ratio of summer‐to‐winter notifications for each year in the study. A P‐value of <0.05 was used for statistical significance. Results Nationally, 18 933 notifications were reported during summer months from December to February 2005‐2016. There have been increasing summer notifications over time, which corresponded to similarly increased notifications in winter. A significant upward trend was observed for rate of notification during summer period over these years, P < 0.01. However, the ratio of summer‐to‐winter notifications demonstrated that while notifications have increased, the ratio has not increased markedly over the years and did not show a significant trend. No seasonal trend in rates of hospitalisation for influenza and pneumonia, respiratory and circulatory diagnosis was observed over the studied years. Conclusion This study provides a clearer understanding of the epidemiology and burden of inter‐seasonal influenza and trends over time in Australia. The ratio of summer‐to‐winter notifications remains relatively constant and is supported by reasonably constant hospitalisation rates over the years.


| BACKG ROU N D
Influenza is an epidemic infection which affects millions of people around the world annually. Globally each year, the incidence of human influenza rises and falls in temperate regions during winter. 1 This is known as seasonality and is largely believed to be driven by favourable climate conditions for viral survival and transmission, 2 and changes in population behaviours such as crowding. 3 In tropical and subtropical regions, however, this distinct seasonal pattern is less clearly defined 4 with cases occurring year round, particularly in tropical East and South-East Asia. 5 Inter-seasonal notifications of influenza, in the summer months, were rare a decade ago. 6 However, notifications of influenza appear to have increased in summer and outside of the usual influenza season in Australia. 7 Australia is a large continental land, which spans many climatic zones, from tropical and subtropical, to temperate and cool temperate. In temperate and cool temperate Australia, the seasonal rise in influenza incidence typically begins around May and falls by October. 8 Cases that occur outside this period are characterised as inter-seasonal migrant viruses-imported from the opposing hemisphere or the tropics. 9 Travel has increased exponentially over the last two decades, with about 766 600 short-term visitor arrivals to Australia during July 2018, an increase of approximately 43% when compared with records within the last two decades. 10 Inter-seasonal cases are often sporadic and thought to be unviable for ongoing transmission, and hence, their lineages become extinct. 11 However, a recent study has shown that in some instances, lineages of inter-seasonal influenza in Australia have continued to circulate into the typical winter season. 7 Furthermore, in some cases, seasonal lineages of influenza B demonstrated persistent transmission across the inter-seasonal period, effectively bridging one season to another. If inter-seasonal lineages can be shown to persist into and across typical seasonal periods, this gives rise to questions of seasonal onset. One recent study has investigated the onset of seasonal influenza across each climatic zone in Australia. They showed the onset of epidemics varies season to season; however, there is remarkable synchronisation of epidemic onset across Australia, with a winter peak typically in August, despite the significant distance involved and variation in climate. 12 There have also been reported increased cases of summer influenza or out-of-season epidemic in other countries. 13 We aimed to determine whether inter-seasonal influenza activity has been increasing in Australia over time.  Every year, influenza seasonal activity occurs between May and October in Australia, with a typical peak in August. 16 Inter-seasonality was defined as influenza activity occurring in the period from December to February and was compared to influenza activity during the winter months (June-August). The NNDSS data, which are validated and confirmed as influenza, were analysed to examine an association between inter-seasonality of influenza and the impact on morbidity. The 2009 pandemic year was excluded from the analysis, which was restricted to seasonal influenza in the study.

| ME THODS
Routine data sources including national hospitalisations data from the Department of Health, AIHW, 17 and population data from the Australian Bureau of Statistics (ABS), Australian Government were also used in the study. 18 Weekly hospitalisations data were extracted for three disease categories, using the clinical diagnosis coding ICD-10-AM with regard to the principal diagnosis of influenza and pneumonia (J09-J18), all respiratory (J00-J99) and circulatory disease (I00-I99) for all age group. Data were then aggregated and analysed as monthly hospitalisations in the study. We used monthly rate of hospitalisation per 100 000 population in each diagnosis category for summer period from December to February, and for winter from July to August, between 2005 and 2016. We tested whether there was a seasonal trend in the rate of hospitalisation during summer.

| Analysis methods
A descriptive analysis was conducted using the notification and hospitalisation data by year. One of the reasons for increased summer notifications is increased testing for influenza over time. To adjust for changes in testing over time, we calculated an adjusted ratio (ratio of summer-to-winter notifications) for each year assuming that a real increase in summer cases may be reflected in an increased ratio over time. We tested for trend using the seasonal Mann-Kendall test to detect trends over time during the studied years using xlstat software. 19 IBM spss Statistics, Version 22 was used, 20 and a P-value of <0.05 was applied for the statistical significance.  Table 1 shows the number of laboratory-confirmed influenza notifications and notification rate per 100 000 population during summer period  (Table 1). We tested notifications for trend over time during the summer months using the Seasonal Mann-Kendall test, and the results showed that there was a trend to increased summer notifications, which was statistically significant (P < 0.01). Similarly, a significant trend was also observed for notification rate during the summer period over the years, P < 0.01. Table 1 indicates that while the notifications are increased, the ratio of summer-to-winter notifications or the adjusted ratio has not increased markedly. Besides, testing of the adjusted ratio for seasonal Mann-Kendall test over the years did not show a significant trend, P = 0.13.

| RE SULTS
We analysed summer notifications across all states and territories in Australia over the study period. Figure 1  population per year, respectively. A statistically significant trend in summer notification rates was also observed in all states and territories over the years (Table 2).
A descriptive summary of hospitalisation rates across three diagnosis categories over the studied years is shown in Table 3. A mean hospitalisation rate of 64.2/100 000 population in influenza and pneumonia to a mean rate of 509.9/100 000 population in circulatory hospitalisations per year were observed over the years. There was no significant trend in increased summer rates of hospitalisations for all diagnosis categories during the studied years (Table 3).

| D ISCUSS I ON
This study provides a clearer understanding of the epidemiology and burden of inter-seasonal influenza in Australia and trends over time.
TA B L E 1 Laboratory-confirmed influenza notifications and notification rate per 100 000 population, and adjusted ratio during summer and winter months,  The state-based data showed varied notification rates as well as significant increasing trend in summer notification rates across the states over time. Another Australian study found that there was marked increase in influenza notifications in Australia during summer period of 2010-2011; however, authors reported that those high notifications were simply due to variation in virus circulation during summer and no association or impact had been reported with regard to increased cases. 21 Similar to our study, evidence of increased notifications was reported in Australia after the pandemic. 21,22 Testing for influenza may have increased after the pandemic due to increased awareness of the disease and the disease impact in the society. In addition, availability and use of rapid point of care tests for influenza have also increased since 2009 after the provision of public funding for rapid testing by the government. 22 Some countries experience more than one peak of influenza To conclude, we found increasing summer notifications over time, which corresponded to similar increased notifications in winter, without changes in influenza hospitalisations over the same period.
This supports the conclusion that the apparent increase in incidence of influenza during summer is a result of increased testing over time.
These results should be interpreted carefully in conjunction with other surveillance systems to understand the overall epidemiology of influenza infection in the country.

ACK N OWLED G EM ENTS
We acknowledge the data sources: the Department of Health,